School Psychology © 2019 American Psychological Association 2578-4218/19/$12.00 2019, Vol. 34, No. 4, 376 –385 http://dx.doi.org/10.1037/spq0000282 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Parent Involvement, Approaches to Learning, and Student Achievement: Examining Longitudinal Mediation Christopher James Anthony Julia Ogg Oklahoma State University Northern Illinois University Although there is evidence to suggest that parent involvement (PI) in children’s education positively impacts their academic success, the mechanisms of this effect are less well studied. One potential mechanism is a set of student-level motivational and behavioral factors labeled approaches to learning (ATL). The purpose of the current study was to utilize rigorous longitudinal methodology to evaluate whether ATL mediate the relationship between PI and student academic achievement. Using a large sample drawn from the Early Childhood Longitudinal Study—Kindergarten Cohort (ECLS-K), three sets of analyses were conducted focusing on three different types of PI (home-based involvement, schoolbased involvement, and home–school communication). Longitudinal mediation analyses indicated that only school-based involvement and home–school communication predicted student reading achievement and that this relationship was only mediated by ATL for school-based involvement. These findings contribute to the literature base on PI and represent a methodological advance to addressing these important mediational questions. Impact and Implications The current study utilized rigorous methodology to examine whether approaches to learning, a variable including student motivation, persistence, and engagement, might help explain how parent involvement affects student achievement. Results indicated that approaches to learning explained this relationship partially for one type of parent involvement (school-based involvement). Additionally, the findings shed light on inconsistencies in prior research and the study demonstrated the use of appropriate methodology for answering these important questions. Keywords: parent involvement, home-based involvement, school-based involvement, home–school communication, approaches to learning Supplemental materials: http://dx.doi.org/10.1037/spq0000282.supp ATL, which are learning-related attitudes and behaviors, represent an important plausible mechanism by which PI might influence student achievement. Indeed, there are theoretical models (Hoover-Dempsey et al., 2005) and initial empirical evidence supporting this pattern (Xu, Kushner Benson, Mudrey-Camino, & Steiner, 2010). Furthermore, such a relationship might be especially important in kindergarten, when the patterns that form the basis of later PI are first developed. Previous research examining this important mediational question has been cross-sectional and focused on older students (e.g., fifth graders). As such, the goal of the present study was to expand on this research to longitudinally examine whether ATL mediate the relationship between kindergarten PI and later academic achievement. The link between parent involvement (PI) and academic achievement is well studied, and meta-analyses have indicated small to moderate correlations between these constructs (X. Fan & Chen, 2001; Jeynes, 2005). Yet the mechanisms by which PI may lead to enhanced academic achievement are less well understood. Recently, increasing scholarly attention (e.g., Anthony, DiPerna, & Amato, 2014) has focused on the role of a set of variables referred to as approaches to learning (ATL) that have been linked with both PI (e.g., Fantuzzo, McWayne, Perry, & Childs, 2004; McWayne, Hampton, Fantuzzo, Cohen, & Sekino, 2004) and academic achievement (e.g., Li-Grining, Votruba-Drzal, Maldonado-Carreño, & Haas, 2010). Parent Involvement and Achievement Christopher James Anthony, School of Teaching, Learning and Educational Sciences, Oklahoma State University; Julia Ogg, Department of Psychology, Northern Illinois University. Correspondence concerning this article should be addressed to Christopher James Anthony, who is now at School of Special Education, School Psychology, and Early Childhood Studies, University of Florida, 2-173 Norman Hall, Gainesville, FL 32611. E-mail: canthony@coe.ufl.edu Although the link between PI and academic achievement is well established (Garbacz, Herman, Thompson, & Reinke, 2017; Jeynes, 2005, 2012), the strength of the association depends on the specific form of PI being investigated. Specifically, theoretical models of PI have recognized the distinction between three types of PI: home-based involvement (HBI), school-based involvement 376 This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PARENT INVOLVEMENT, ATL, AND ACHIEVEMENT (SBI), and home–school communication (HSC; Epstein, 1995; Hoover-Dempsey et al., 2005). Empirical evidence supports this distinction through factor analysis (Fantuzzo, Tighe, & Perry, 1999; Manz, Fantuzzo, & Power, 2004) and through findings suggesting that each form of PI relates differently to outcomes (Fantuzzo et al., 2004; Garbacz, McDowall, Schaughency, Sheridan, & Welch, 2015). HBI includes reading or engaging in other academic activities with children as well as more general intellectual activities such as attending educational events or visiting sites in the community (e.g., a zoo). Meta-analyses have indicated that HBI shows a small to moderate positive association with academic achievement, including broad indicators such as GPA and subject-specific indicators (e.g., reading or math scores) of academic achievement in school-age youth (X. Fan & Chen, 2001; Jeynes, 2005). SBI includes activities in which a parent participates at school (e.g., PTA meetings, family nights). SBI has shown a small, positive association with broad and subject-specific academic indicators in meta-analyses (X. Fan & Chen, 2001; Jeynes, 2005). Finally, HSC involves contact between a child’s parents and teacher or other school personnel (e.g., in-person or via e-mail). This form of involvement has been linked inconsistently to academic outcomes. For example, Fantuzzo et al. (2004) found that although preschoollevel HSC in the fall was positively correlated with spring receptive vocabulary scores, controlling for HBI rendered this association nonsignificant. In another study with a large, national sample of fifth grade students, Xu et al. (2010) found that some aspects of HSC (e.g., homework help) were negatively associated with reading assessment scores. One possibility for this inconsistency is that HSC may concern both positive and negative child behaviors (i.e., HSC may increase in response to increased incidence of problematic behavior). Together, research suggests that at least two of the three forms of PI (HBI and SBI) are positively related to academic achievement, whereas this association is less clear with regard to HSC, highlighting the importance of considering PI as multidimensional. Parent Involvement and Approaches to Learning PI has also been linked with aspects of ATL (W. Fan & Williams, 2010; Fantuzzo et al., 2004; Xu et al., 2010). ATL refer to a broad set of skills, attitudes, and behaviors, typically including motivation, engagement, and other learning-related behaviors (e.g., study skills; DiPerna, Volpe, & Elliott, 2002, 2005). Although many behaviors fall within this domain, ATL are considered to broadly reflect the application of self-regulation or “the ability to manage one’s behavior, emotions, and attention in voluntary and adaptive ways” (Li-Grining et al., 2010, p. 1062) toward academic-related goals. Regardless of which specific variables are included, ATL have consistently been linked with academic achievement (e.g., Anthony et al., 2014). Furthermore, drawing on both ecological theory and social– cognitive theories, PI is linked to ATL in several ways. From an ecological perspective, PI represents mesosystemic contexts (i.e., interactions between a child’s home and school microsystems) hypothesized to impact studentlevel socioemotional, academic, and behavioral functioning (Bempechat & Shernoff, 2012; Bronfenbrenner & Ceci, 1994). Furthermore, PI has been hypothesized to communicate the value of 377 education from parents to children and may also model and scaffold ATL for children (Bempechat & Shernoff, 2012). Empirical research has supported this theoretical connection. Fantuzzo et al. (2004) found that all three forms of PI (SBI, HBI, and HSC) measured in the fall were associated with three forms of ATL measured in the spring of preschool: competence motivation, attention/persistence, and attitudes toward learning. Specifically, competence motivation was positively associated with all three forms of PI, whereas HBI and SBI only were associated with attention/persistence and attitudes toward learning in children enrolled in Head Start. HBI (.30 ⬍ r ⬍ .36) was more strongly associated with all three ATL variables than SBI (.23 ⬍ r ⬍ .25) or HSC (.12 ⬍ r ⬍ .23). In a national sample of fifth grade students, Xu et al. (2010) found that SBI, parental education expectations, homework frequency, and extracurricular activities positively predicted students’ ATL. In the Xu et al. study, two other forms of PI, TV rules and homework help, negatively predicted ATL. These finding indicate that although ATL is associated with all three forms of PI, different types of PI relate differently in terms of strength and direction of association to ATL. Approaches to Learning and Achievement The link between PI and ATL is particularly important because a number of studies have documented a positive association between ATL and academic achievement (Claessens, Duncan, & Engel, 2009; DiPerna, Volpe, & Elliott, 2002; Duncan et al., 2007). ATL-related constructs have been proposed to be one component of academic competence (DiPerna & Elliott, 2002), which also consists of academic skills. This model of academic competence posits that both academic skills and ATL (referred to as “academic enablers” by DiPerna & Elliott, 2002) are integral aspects of students’ academic development and school success. Several studies have supported this conceptualization. For example, in two studies with kindergarten through sixth grade students, DiPerna and colleagues (2002; DiPerna, Volpe, & Elliott, 2005) found a moderate to strong association between ATL constructs, including engagement, motivation, and achievement, in both reading and mathematics. Also, after combining data from six large, longitudinal data sets, Duncan and colleagues (2007) found that a variable that included ATL at school entry had a small positive effect on later reading and math achievement. Such findings are consistent with decades of research tying ATL-related variables to academic achievement (e.g., Farrington et al., 2012; Wang, Haertel, & Walberg, 1993). Parent Involvement, Approaches to Learning, and Achievement Although PI may directly influence achievement (e.g., skill building through homework), and much of the current literature has examined this direct relationship, PI could also indirectly influence achievement through the development of students’ ATL (Bempechat & Shernoff, 2012). Several authors (Grolnick & Slowiaczek, 1994; Xu et al., 2010) have explicitly proposed this hypothesis, and theoretical models of PI have also emphasized this possibility. Specifically, Hoover-Dempsey et al.’s (2005) model of PI proposes that the effects of PI on achievement are mediated through student attributes (i.e., academic self-efficacy, motivation, self-regulation) consistent with ATL. ANTHONY AND OGG This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 378 In an empirical test of ATL as mediators, Grolnick and Slowiaczek (1994) examined whether the impact of PI (including HBI, SBI, and HSC) on school performance was mediated through children’s motivational resources (self-regulation, perceived competence, and control understanding) in a cross-sectional study of middle-school students. Their findings indicated that students’ perceived competence and control understanding (i.e., their knowledge of how outcomes are linked to their behavior) both mediated the relationship between HBI and HSC and students’ grades (averaged across curricular areas). Furthermore, using a national cross-sectional sample of fifth graders from the Early Childhood Longitudinal Study—Kindergarten Cohort of 1998 –1999 (ECLS-K; Tourangeau, Nord, Lê, Sorongon, & Najarian, 2009), Xu et al. (2010) found that ATL mediated the relationship between PI and reading achievement. These findings point to the potential importance of ATL as mediators of the impact of PI on student achievement. Developmental Context of Mediation One important aspect of studying associations between PI and academic achievement is considering the developmental processes in which effects are embedded. Students’ academic skills, as well as their ATL (e.g., Mahatmya, Lohman, Matjasko, & Farb, 2012), develop throughout their academic careers. Furthermore, prior research has found that that levels of PI also differ across grade level. For example, a recent examination of patterns of the three forms of PI (HBI, SBI, and HSC) in a large, multigrade, cross-sectional elementary school sample indicated that levels of SBI were roughly equal across grades, whereas HBI and HSC were lower at the higher elementary school grades (Garbacz et al., 2015). Thus, interrelationships between PI, ATL, and achievement occur within dynamic developmental systems. As such, mediation may vary substantially depending on when and for how long it is examined. One particularly important period may be in kindergarten, when a large majority of parents are interested in becoming involved (McIntyre, Eckert, Fiese, DiGennaro, & Wildenger, 2007). As further indication of the potential importance of examining PI in kindergarten, research has shown that levels of PI in early elementary school are predictive of PI later in elementary school (Englund, Luckner, Whaley, & Egeland, 2004). Thus, interactions between home and school during this period may set the stage for ongoing interrelationships between PI, ATL, and achievement. Such a possibility has not been previously examined, as prior studies have focused on potential mediation in fifth grade (Xu et al., 2010) and in middle school (Grolnick & Slowiaczek, 1994). Current Study The current study addresses several theoretical and methodological gaps in research on PI. A key limitation in research on PI is the use of a wide variety of definitions for PI that are not aligned with the constructs identified through theoretical work (i.e., SBI, HBI, HSC), which may contribute to mixed findings (e.g., some studies combine HSC and SBI). To address this gap, the current study used a consistently empirically supported theoretical model of PI by defining PI with three separate factors: HBI, SBI, and HSC (Fantuzzo et al., 2004; Hoover-Dempsey et al., 2005). Furthermore, much of the prior research on PI, in general, and this mediational question, in particular, has been cross-sectional. This makes it difficult to fully understand the dynamic relation- ships between PI and other variables that occur across time, especially mediational relationships. In fact, many methodologists (e.g., Selig & Preacher, 2009) contend that only longitudinal approaches can appropriately assess mediation, which necessarily occurs over time. To address this shortcoming, the current study builds on prior research by examining mediation that was explicitly longitudinal. This conceptualization also ties in with the importance of considering the dynamic developmental context in which mediation occurs. As such, the current study incorporates both methodological and conceptual strengths by considering mediation longitudinally as embedded within a particular developmental period in children’s school careers. Research Questions and Hypotheses The following broad research question guided the current study: Do ATL mediate the relationship between kindergarten-level PI and later academic achievement? This research question was divided in two subquestions framed according to our methodology: 1. Does kindergarten-level PI (HBI, SBI, and HSC) exert a total effect on later reading achievement? Based on prior research demonstrating a positive association between HBI and SBI (e.g., X. Fan & Chen, 2001), we hypothesize that both HBI and SBI will have a small total effect on reading achievement. Given the lack of consistency between HSC and achievement (e.g., Fantuzzo et al., 2004), we hypothesize no significant total effect between HSC and reading achievement. 2. If a total effect is evident, does kindergarten-level PI (HBI, SBI, and HSC) exert an indirect effect (indicating mediation) on later reading achievement through ATL? Based on prior research in this area, we hypothesize that ATL will mediate the relationship between HBI and SBI (Grolnick & Slowiaczek, 1994; Xu et al., 2010) and reading achievement. Because we hypothesize no significant total effect between HSC and reading achievement, we also hypothesize no significant indirect effect. We also explored the individual components of the models used to evaluate longitudinal mediation (i.e., PI on ATL and ATL and reading achievement). Overall, we hypothesize that HBI and SBI will be positively associated with ATL, whereas HSC will not be associated with ATL. In addition, we anticipate a small to moderate positive association between ATL and reading achievement based on prior research in this area (e.g., Claessens et al., 2009; Duncan et al., 2007). Method Participants Participants for this study were drawn from the ECLS-K, which includes a nationally representative sample followed from kindergarten through the eighth grade. Based on data availability, data were drawn from three waves for HBI (during the fall of the kindergarten and spring of the first and third grade waves) and HSC (during the spring of the kindergarten, first, and third grade waves) analyses, and from four waves for SBI analyses (during the spring of the kindergarten, first, third, and fifth grade waves). PARENT INVOLVEMENT, ATL, AND ACHIEVEMENT Because of the complex sampling design utilized by the ECLS-K, sample weights are needed to ensure that statistical estimates are representative of the target population. We identified and applied the appropriate sampling weight for each set of analyses. Prior to analysis, cases were dropped if they were missing data on the weight variable. An additional 11 cases were missing all analytic variables for the SBI data and were dropped. After dropping these cases, the initial sample was 16,555 for both the HSC and HBI samples, and 12,011 for the SBI sample (see Table 1). This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Measures Parent involvement. PI was conceptualized as multidimensional and measured in three ways for the three different models in this study. First, HBI was indicated by nine parent-rated items utilizing a Likert scale ranging from 1 (not at all) to 4 (every day). SBI was indicated by five parent-rated dichotomous (yes–no) items. Finally, HSC was indicated by four teacher-rated dichotomous (yes–no) items. These sets of items have been used in previous studies (e.g., Galindo & Sheldon, 2012) to indicate PI and have been linked as expected with various outcomes (e.g., Xu et al., 2010; Youn, Leon, & Lee, 2012). Approaches to learning. ATL was assessed through a subscale of the social rating scale developed specifically for the ECLS-K. This subscale included six items measuring teacher-rated learning behaviors, including behaviors related to working independently, eagerness to learn, and student persistence. These items Table 1 Weighted Demographic Characteristics of Participants for Home-Based Involvement, School-Based Involvement, and Home–School Communication Samples Home-based School-based Home–school involvement involvement communication (n ⫽ 16,555) (n ⫽ 12,011) (n ⫽ 16,555) Characteristic Weighted % Weighted % Weighted % Girls Race/ethnicity White, non-Hispanic African American Hispanic Asian Other Region Northeast Midwest South West Area Central city Urban-fringe/largetown Small town and rural School type Public school Private school Disability status Child with disability Child without disability 49 49 49 58 16 19 3 5 58 16 19 3 5 58 16 19 3 5 18 23 37 22 18 23 37 22 18 23 37 22 37 42 37 42 37 42 21 21 21 85 15 85 13 85 15 16 84 16 84 16 84 Note. Percentages do not sum to 100 in some cases because of rounding. Region, area, school type, and disability status reported for Wave 1 (kindergarten) in each model. 379 were rated on a Likert scale from 1 (never) to 4 (very often). This scale has been used extensively in previous investigations using the ECLS-K (e.g., Claessens et al., 2009; Li-Grining et al., 2010) and has been found to relate as expected with academic achievement in both reading and mathematics (e.g., Anthony et al., 2014). Student achievement. Student achievement was measured through the directly administered reading tests developed using item response theory (IRT) methodology for the ECLS-K. We selected reading as the primary outcome measure for this study because the ability to comprehend text is fundamentally important for achievement across broad content domains throughout students’ educational careers (Ardoin & January, 2018).1 These assessments consisted of various questions focused on reading domains including vocabulary and passage comprehension. Reliability of theta (an IRT-based internal consistency index) ranged from .91 to .96 for these scores across waves. Furthermore, as validity evidence, the patterns of reading score growth across the ECLS-K sample frame were similar to those observed in similar longitudinal studies across the same period (Najarian, Pollack, & Sorongon, 2009). Descriptive information for student achievement by wave is shown in Table 2. Covariates. Several covariates were included in all models, including student gender, family socioeconomic status (SES; a composite derived from measures of income, educational status, and occupational status), and family structure. Gender and SES were time invariant covariates, whereas family structure (parents’ reported marital status) was included as a time-varying covariate. Descriptive statistics for covariates are displayed in Table 2. Procedures Parent data were collected primarily through telephone interviews conducted by trained ECLS-K staff. Child data consisted of directly administered assessment batteries also conducted by trained ECLS-K staff. Finally, teacher data were collected through self-administered questionnaires. Further details regarding data collection can be found in the ECLS-K methodology reports (e.g., Tourangeau, Lê, & Nord, 2005). Institutional review board approval for analysis of these data as an extant dataset was obtained through the first author’s institution. Data Analysis Missing data. There were varying levels of missing data across samples, ranging from 0% to 31% (median ⫽ 15%) across HBI variables, from 0% to 21% (median ⫽ 9%) across SBI variables, and from 0% to 52% (median ⫽ 19%) across HSC variables. Roughly 59%, 47%, and 87% of cases were missing at least one data point across HBI, SBI, and HSC models, respectively. Because we used the mean and variance adjusted weighted least squares estimator (WLSMV), which relies on pairwise deletion, we used multiple imputation prior to analyses to appropriately address missing data. Based on the recommendation to generate more imputed data sets than the number of incomplete cases (White, Royston, & Wood, 2011), we generated 60, 50, and 90 imputed data sets for the HBI, SBI, and HSC models, respectively. 1 Although we have focused on reading in the current study, we also recognize the importance of math skills in elementary school. Thus, we have run parallel models for math. These results can be seen in Tables 1–2 in the online supplemental materials. ANTHONY AND OGG 380 Table 2 Weighted Descriptive Statistics for Analysis Variables Across Wave Home-based involvement This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Variable Reading achievement Wave 1 Wave 2 Wave 3 Wave 4 Wave 1 SES Married parents Wave 1 Wave 2 Wave 3 Wave 4 School-based involvement Home–school communication M SD M SD M SD 35.20 77.22 126.27 — ⫺.03 9.98 23.83 28.12 — .79 46.11 77.12 126.27 148.68 ⫺.03 13.60 23.87 28.28 26.56 .79 46.25 77.15 126.20 — ⫺.04 13.79 23.85 28.11 — .78 69 69 69 — — — — — — — — — 68 68 69 — — — — — 69 68 69 67 Note. Married parents reported as percentages. Reading achievement scores are on a scale ranging from 0 to 212. Socioeconomic status (SES) scores are standardized with a mean of 0 and standard deviation of 1. Latent variable approach. Rather than including PI and ATL variables in models as sum scores, we utilized a latent variable modeling approach in which each construct (HBI, SBI, HSC, and ATL) was indicated by several items. We favored this approach for two reasons. First, latent variable approaches address measurement error and disattenuate relationships between constructs (Cole & Preacher, 2014). Next, latent variable approaches allow for explicitly testing measurement invariance, an important assumption of all longitudinal models (Little, 2013). Because the availability of PI measures differed across waves, we conducted three separate mediation models for HBI, SBI, and HSC variables. Analysis steps. Several steps based on Cole and Maxwell (2003) and Little (2013) were conducted to test our hypotheses. All models were estimated using Mplus Version 7 (Muthén & Muthén, 2017). Because of the categorical nature of the data, the WLSMV Parent Involvement Kindergarten Approaches to Learning Kindergarten Reading Kindergarten arPI1 arATL1 arREAD1 Parent Involvement 1st Grade Approaches to Learning 1st Grade Reading 1st Grade arPI2 arATL2 arREAD2 estimator was utilized for all analyses. Overall model fit was evaluated relative to Hu and Bentler’s (1999) criteria. Specifically, because chi-square values are sensitive to sample size (Little, 2013), we emphasized root mean squared error of approximation (RMSEA) values less than .06 and comparative fit index (CFI)/ Tucker-Lewis index (TLI) values greater than .95 when interpreting overall model fit. Relative model fit was evaluated based on the ⌬CFI ⬍ .01 criteria recommended by Cheung and Rensvold (2002; i.e., if a more restrictive model did not result in the CFI decreasing by .01 or more, the more restrictive model was retained). Because the default null model used to calculate CFI and TLI is not appropriate for longitudinal models, a modified null model was conducted and used to calculate all CFI and TLI statistics (Little, 2013). Step 1: Test measurement invariance. Following steps based on Little (2013), a series of increasingly restrictive models were tested to evaluate whether constructs were equivalent and measured equivalently across waves. First, a configural model was tested in which loadings were freely estimated for constructs across time. Assuming the configural model evidenced adequate fit, a metric invariance model was tested in which factor loadings were fixed to be equivalent across waves. If the metric model did not lead to unacceptable degradation of model fit relative to the configural model, a scalar model was tested in which all thresholds were fixed to be equivalent across time. Once scalar measurement invariance was supported, the constraints of this model were imposed on all subsequent models. Step 2: Fit a saturated model. Once scalar invariance was supported, we fit a saturated panel model in which variables were regressed on all analytic variables from all previous waves. Statistical significance of paths from this model were evaluated to help determine the final model for each analysis. Step 3: Fit and test a final model. In the last step of the analyses, a model including all paths consistent with longitudinal partial mediation (i.e., those in Figure 1) and all statistically significant paths from Step 2 was tested. Such a model was considered a pruned and final model, assuming it did not lead to Parent Involvement 3rd Grade Approaches to Learning 3rd Grade Reading 3rd Grade arPI3 arATL3 arREAD3 Parent Involvement 5th Grade Approaches to Learning 5th Grade Reading 5th Grade Figure 1. Example of longitudinal mediation model including paths estimated for all mediation models. The a, b, and c paths are labeled consistent with standard mediation models (e.g., Baron & Kenny, 1986). ar ⫽ autoregressive; PI ⫽ Parent Involvement; ATL ⫽ Approaches to Learning; READ ⫽ Reading Achievement. PARENT INVOLVEMENT, ATL, AND ACHIEVEMENT This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. substantial degradation in model fit (as evaluated by ⌬CFI value relative to the scalar invariance model). Once this model was identified, total, direct, and indirect effects were calculated by summing the products of different paths constituting different types of effects. Direct effects were calculated by summing the products of all paths by which PI in the first wave of the study affected reading achievement in the final wave of the study without passing through ATL. Indirect effects represented the sum of the products of all paths by which PI in the first wave affected reading achievement in the final wave passing through ATL. For example, the following equations would represent direct and indirect effects in Figure 1: Direct effect ⫽ (c1 * arREAD3) ⫹ (arPI1 * c2) and Indirect effect ⫽ (a1 * arATL2 * b2) ⫹ (a1 * b1 * arREAD3) ⫹ (arPI1 * a2 * b2). Total effects represent the sum of all direct and indirect paths. Statistical significance of these effects were evaluated using the Monte Carlo method outlined by Preacher and Selig (2012). These calculations were conducted with the monteCarloMed option from the semTools package (semTools Contributors, 2016) in R Version 3.3.1 (R Core Team, 2016). If the generated 95% confidence interval (CI) included zero, the effect was not considered statistically significant. To gauge magnitude of statistically significant effects, direct, indirect, and total effects were recalculated with fully standardized paths. These standardized effects were interpreted according to Cohen’s (1988) criterion for small (⬍ .10), medium (.10 –.30), and large (⬎ .30) effects. Results Home-Based Involvement The configural model fit the data adequately (see Table 3), and neither the metric nor the scalar models led to unacceptable decrements in model fit (⌬CFI ⫽ ⬍.001 and .009 for these models, 381 respectively). Fit information from the final model did not indicate substantial decrements in model fit (⌬CFI ⫽ .011) relative to the scalar invariance model. Constrained unstandardized loadings ranged from .40 to .58 (median ⫽ .54) for HBI indicators and from .68 to .89 (median ⫽ .82; Table 4) across ATL indicators. Based on parameter estimates from the final model, neither the total effect of HBI on ATL (total effect ⫽ 0.005, 95% CI [⫺0.007, 0.016]) nor the direct effect of HBI on reading (direct effect ⫽ 0.004, 95% CI [⫺0.011, 0.011]), nor the indirect effect of HBI on reading through ATL (indirect effect ⫽ 0.001; 95% CI [⫺0.001, 0.002]) was statistically significant. Standardized parameter estimates indicated that HBI did not have a statistically significant effect on either ATL or reading in any wave. Likewise, ATL did not have a statistically significant effect on HBI in any wave. In contrast, ATL had a positive effect on reading both from kindergarten to first grade ( ⫽ .16, p ⬍ .001) and third grade ( ⫽ .09, p ⬍ .001), and from first grade to third grade ( ⫽ .11, p ⬍ .001). Reading scores in kindergarten had a negative effect on HBI in both first grade ( ⫽ ⫺.09, p ⬍ .001) and third grade ( ⫽ ⫺.05, p ⫽ .047). Reading scores in first grade also had a negative effect on HBI in third grade ( ⫽ ⫺.06, p ⫽ .022). Reading scores in kindergarten had a positive effect on ATL in first grade ( ⫽ .18, p ⬍ .001), and reading scores in first grade had a positive effect on ATL in third grade ( ⫽ .13, p ⬍ .001). School-Based Involvement For the SBI models, the configural invariance model fit the data adequately (see Table 3). Neither the metric nor the scalar models led to substantial decrements in model fit (⌬CFI ⫽ .001 for both models), and thus scalar invariance was imposed on all subsequent models. Constrained unstandardized loadings ranged from .35 to .59 (median ⫽ .47; Table 4) for SBI indicators and from .68 to .90 (median ⫽ .82) across ATL indicators. After fixing all nonsignificant paths from the full model to zero, a further two paths were also nonsignificant and constrained to zero in the final model. The final model did not lead to substantial decrements in model fit Table 3 Model Fit Statistics Model Home-based involvement Configural invariance Metric invariance Scalar invariance Final model School-based involvement Configural invariance Metric invariance Scalar invariance Final model Home–school communication Configural invariance Metric invariance Scalar invariance Final model 2(df) RMSEA CFI ⌬CFI TLI 3,797.45 (1158) 3,814.63 (1184) 4,867.75 (1270) 3,625.18 (1302) .014 .014 .016 .013 .977 .977 .968 .979 — ⬍.001 .009 .011 .972 .972 .965 .978 1,928.52 (1132) 1,887.81 (1159) 2,054.06 (1222) 2,075.20 (1290) .009 .008 .009 .008 .991 .992 .991 .991 — .001 .001 ⬍.001 .989 .990 .989 .990 3,205.95 (528) 2,902.37 (544) 2,966.77 (584) 2,891.69 (616) .020 .019 .018 .017 .981 .984 .983 .984 — .003 .001 .001 .975 .979 .980 .982 Note. All chi-square statistics were statistically significant. All statistics are averages over 60, 50, and 90 imputed data sets for home-based involvement, school-based involvement, and home–school communication models respectively. df ⫽ degrees of freedom; RMSEA ⫽ root mean squared error of approximation; CFI ⫽ comparative fit index; TLI ⫽ Tucker-Lewis index. ANTHONY AND OGG 382 Table 4 Unstandardized Factor Loadings for Final Measurement Invariance Models This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. Item Home-based involvement Tell stories? Sing songs? Help to do arts and crafts? Involve in household chores? Play games or do puzzles? Talk about nature or do science projects? Build something or play with construction toys? Play a sport or exercise together? Read books? School-based involvement Attended an open house or back-to-school night? Attended a meeting of a PTA, PTO, or Parent–Teacher Organization? Attended a school or class event? Volunteered at the school or served on a committee? Participated in fundraising for your child’s school? Home–school communication Attended regularly scheduled conferences at your school? Attended informal meetings that you initiated to talk about the child’s progress? Returned your telephone calls? Initiated contact with you? Approaches to learning Keeps belongings organized? Shows eagerness to learn new things? Works independently? Easily adapts to changes in routine? Persists in completing tasks? Pays attention well? Loading .58 .49 .40 .54 .54 .55 .55 .48 .50 .51 .35 .47 .59 .39 .85 .86 .90 .25 .69 .77 .89 .68 .86 .89 Note. Unstandardized loadings presented, as these are constrained to be equal across time. Approaches to learning loadings presented from homebased involvement model, but loadings from other models were very similar (all differences in loadings ⱕ.02). PTA ⫽ Parent-Teacher Association; PTO ⫽ Parent-Teacher Organization. relative to the scalar model (⌬CFI ⬍ .001). Based on parameter estimates from the final model, the total effect of SBI in kindergarten on reading achievement was statistically significant (total effect ⫽ 0.11, 95% CI [0.09, 0.13], standardized effect ⫽ 0.26). Next, the direct effect of SBI in kindergarten on reading in the fifth grade was also statistically significant (direct effect ⫽ 0.10, 95% CI [0.08, 0.12], standardized effect ⫽ .25), as was the indirect effect of SBI on reading through ATL (indirect effect ⫽ 0.004, 95% CI [0.002, 0.007], standardized effect ⫽ 0.01). Standardized parameter estimates indicated that SBI had a positive effect on ATL from kindergarten to first grade ( ⫽ .07, p ⫽ .042), and from first grade to third grade ( ⫽ .07, p ⫽ .011), but not from third grade to fifth grade. SBI in kindergarten also had a positive direct effect on reading in first grade ( ⫽ .06, p ⫽ .003), third grade ( ⫽ .33, p ⬍ .001), and fifth grade ( ⫽ .22, p ⫽ .001). In contrast, SBI in first grade had a negative effect on reading in third grade ( ⫽ ⫺.16, p ⫽ .021), and SBI in third grade had a negative effect on reading in fifth grade ( ⫽ ⫺.10, p ⫽ .004). With regard to ATL, no statistically significant effects of ATL on SBI emerged in the final model. ATL in kindergarten had a positive effect on reading in first and third grades ( ⫽ .14, p ⬍ .001, and  ⫽ .05, p ⫽ .020, respectively). Likewise, ATL had a positive impact on reading from first grade to third grade ( ⫽ .12, p ⬍ .001) and from third grade to fifth grade ( ⫽ .06, p ⬍ .001). Reading scores in kindergarten had a positive effect on SBI in third grade ( ⫽ .07, p ⫽ .026), but no other interval effect reached statistical significance. Finally, reading scores had a positive effect on ATL from kindergarten to first grade ( ⫽ .14, p ⬍ .001) and from first grade to third grade ( ⫽ .06, p ⫽ .011), but not from third grade to fifth grade. Home–School Communication Finally, analysis steps were repeated with the HSC model. First, the configural model evidenced adequate model fit (see Table 3). Neither the metric (⌬CFI ⫽ .003 relative to the configural model) nor the scalar (⌬CFI ⫽ .001 relative to the metric model) models led to substantial decrements in model fit, and thus scalar invariance was supported. Constrained unstandardized loadings ranged from .25 to .90 (median ⫽ .86; Table 4) for HSC indicators and from .69 to .90 (median ⫽ .84) across ATL indicators. The final model fit the data well (see Table 3) and was associated with a low ⌬CFI value (.001 relative to the scalar model). Based on parameter estimates from the final model, the total effect of HSC on reading achievement was statistically significant (total effect ⫽ 0.103, 95% CI [0.084, 0.123], standardized effect ⫽ .213) as was the direct effect of HSC on reading (direct effect ⫽ .102, 95% CI [0.083, 0.120], standardized effect ⫽ .209). The indirect effect; however, was not statistically significant (indirect effect ⫽ 0.002; 95% CI [⫺0.001, 0.004]). With regard to parameter estimates in the final model, standardized estimates indicated that HSC did not have a statistically significant effect on ATL in any wave. In contrast, HSC in kindergarten had a positive effect on reading in both first grade ( ⫽ .10, p ⬍ .001) and third grade ( ⫽ .21, p ⬍ .001), but HSC in first grade had a negative effect on reading in the third grade ( ⫽ ⫺.08, p ⫽ .005). The final model also indicated that ATL in kindergarten had a negative effect on HSC in first grade ( ⫽ ⫺.11, p ⫽ .006), but no effect from first grade to third grade. ATL had a positive effect on reading both from kindergarten to first grade ( ⫽ .11, p ⬍ .001), and from first grade to third grade ( ⫽ .13, p ⬍ .001). Finally, reading scores in kindergarten had a positive effect on HSC in first grade ( ⫽ .19, p ⬍ .001), but not from first grade to third grade. Reading also had a positive effect on ATL both from kindergarten to first grade ( ⫽ .18, p ⬍ .001) and from first grade to third grade ( ⫽ .11, p ⬍ .001).2 Discussion The current study examined whether ATL mediated the relationship between three forms of PI (HBI, SBI, HSC) and reading achievement. Several major findings emerged from our analyses. First, the finding that HBI was not associated with statistically significant total or direct effects on children’s reading was unexpected. From an ecological perspective, HBI is unique relative to 2 As stated previously, we conducted equivalent analyses with the ECLS-K mathematics test. These results indicated that although the general mediational patterns found with reading held with math, the overall magnitude of these effects was slightly less than half as large as the effects observed with reading. These results are available in Tables 1–2 in the online supplemental materials. This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PARENT INVOLVEMENT, ATL, AND ACHIEVEMENT SBI and HSC because it likely involves the least amount of interaction between the home and school microsystems. As such, the coordination between systems may represent a particularly important way for parents to be involved in order to improve school-based outcomes (e.g., achievement). This possibility could have been compounded by the nature of the items used to indicate HBI in this study. Specifically, items for this factor were not focused on school-based academic material (e.g., homework completion), whereas previous investigations of HBI have included some content related to such academic activities (Fantuzzo et al., 2004). It is possible that these different types of HBI (broadly cognitively stimulating vs. academic-focused) have differential effects and should be considered separately. Next, SBI and HSC significantly predicted children’s reading achievement when examined longitudinally. Based on prior research (X. Fan & Chen, 2001; Fantuzzo et al., 2004), we anticipated that SBI would be a significant predictor of achievement. Our findings corroborated this hypothesis using a rigorous, longitudinal design. It is striking that this longitudinal effect persisted and was moderate across several years during children’s early educational experiences, supporting the importance of these forms of PI for student achievement. In contrast to SBI, we predicted the HSC would not predict achievement, considering inconsistent findings linking HSC to achievement (e.g., Fantuzzo et al., 2004). This hypothesis was rejected by the overall model. Instead, our findings indicated that HSC levels in kindergarten positively predicted reading achievement in third grade. Overall, these findings highlight the importance of developing strong PI early in elementary school, as the current study showed that the academic benefits last into later elementary school. Prior research has indicated a majority of parents want to be involved in their child’s transition to kindergarten; thus, school entry appears to be an excellent time to establish home–school relationships (McIntyre et al., 2007). Although not the emphasis of the study, the utilized longitudinal design can potentially shed light on some reasons for the inconsistent patterns found in the HSC literature. In the final HSC model, HSC in kindergarten positively predicted reading in both first and third grade, but HSC in first grade negatively predicted third grade reading. Because kindergarten levels of HSC are already accounted for by virtue of the study design, this negative relationship represents the effect of HSC in first grade on third grade reading controlling for the effect of kindergarten levels of HSC. As such, the effect of HSC in first grade represents the effect of changes in relative levels of HSC over and above initial HSC levels in kindergarten. Thus, the final model indicates that initial kindergarten levels of HSC are consistently positively predictive of reading, but increases in HSC from kindergarten to first grade are negatively predictive of later reading. This distinction could potentially shed light on why findings in this domain have been conflicting (e.g., Izzo, Weissberg, Kasprow, & Fendrich, 1999), as different studies may have captured different aspects of HSC (levels vs. changes in level). For example, in a cross-sectional approach, high levels of measured HSC may be related to (a) high levels of HSC regularly engaged in by parents or (b) recent increases in HSC in response to some student difficulty. These different aspects of HSC could have markedly different effects, as were found in this study, and could help explain previous conflicting results. Interestingly, the same pattern of differential effects was also evident for SBI. Specifically, kindergarten levels of SBI positively predicted reading across all waves, 383 whereas later levels of SBI had negative effects on reading in later waves. Together, these findings suggest an important distinction between levels of HSC and SBI and changes in these variables. Future research should examine these patterns in more detail. Our finding for our main research question was that statistically significant mediation of PI on students’ reading through student ATL was present only for SBI. This finding stands in contrast to previous findings (e.g., Xu et al., 2010) and was likely influenced by the methodology used, as methodologists (e.g., Maxwell, Cole, & Mitchell, 2011) have shown that apparently strong mediational effects estimated with cross-sectional models are often rendered null when examined longitudinally. The general lack of statistically significant mediation appears to be more related to the lack of a relationship between PI and ATL (for both HSC and HBI) than because of the relationship between ATL and reading (which was positive and statistically significant across all intervals and models). PI variables did not exert any effects on ATL for any interval except the small effect of SBI on ATL from kindergarten to first grade and from first grade to third grade. The lack of significant mediation indicates that there are other important mechanisms to explore to better understand the relationship between PI and student achievement. Despite the strengths of this study, there are several important limitations. First, the ECLS-K data were collected starting in 1998 and 1999. It is possible that the dynamics of PI, ATL, and academic achievement have changed since this time. For example, the increase in the use of electronic communication since the time of data collection is significant, particularly in the domain of HSC (Thompson, Mazer, & Flood Grady, 2015). Next, it is possible that mediation was not found because the measure used for ATL was too broad. Researchers often separate constructs included under the ATL umbrella (e.g., motivation, engagement, self-control), and it is possible that PI has effects on some but not all ATL constructs. Likewise, this study focused specifically on reading,3 whereas other subject-specific outcomes (e.g., science), as well as more global indicators of achievement, were not tested. Finally, although teachers have been shown to be good reporters of HSC, it is possible that different patterns would have emerged had parents also reported HSC. This possibility should be explored in future research. It is also important to note that although consideration of the examined relationships in their developmental context was a strength of the study, the results are specific to the measured window. Thus, mediational patterns may vary substantially when examined in other developmental windows, and future research should continue to explore such a possibility with longitudinal data. The transition to kindergarten is a particularly important time to study, as prior research has indicated that parents are interested in being involved in this transition (McIntyre et al., 2007); however, it is also important to understand other important transitions such as the transitions to middle school and high school. Finally, because variables were not measured in equivalent waves, PI variables had to be examined separately rather than in an integrated model. There are various directions for future research. First, future research should continue to explore alternate mechanisms of the relationship between PI and achievement to better understand 3 Supplemental mathematics models, available in Tables 1–2 in the online supplemental materials, yielded roughly equivalent results. ANTHONY AND OGG This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. 384 the effects of PI. Next, future research should continue to refine the multidimensional conceptualization of PI used in this study and explore other important distinctions such as the aforementioned distinction between generic and academic HBI. Finally, the differential effects of levels versus changes in levels in SBI and HSC offer a potential reason for inconsistent effects in the literature. Other methodological approaches (e.g., latent state-trait models; Prenoveau, 2016) decompose variance related to stable and occasion-specific factors and might be profitably applied to more specifically address this question in future research. The current findings also have several implications for practice. First, the overall positive total effects of kindergarten levels of SBI and HSC on reading achievement were moderate, providing support for continuing to view these constructs as important targets for educational interventions. Next, it is important to discuss the distinction between initial levels and changes in levels of HSC (and SBI). Practitioners should not assume that our results indicate that increases in HSC and SBI in and of themselves negatively impact student achievement. Rather, it is likely that the reason that these relationships were found were related to some other variables (e.g., some student difficulty) affecting both the increases in HSC and SBI and reading achievement. Finally, our results indicated that, in general, PI did not positively or negatively affect ATL. These results would suggest that practitioners should consider other potential factors affecting ATL when evaluating children’s ATL in schools, including characteristics of classrooms (e.g., Urdan & Schoenfelder, 2006) and teachers (e.g., Roorda, Koomen, Spilt, & Oort, 2011). Ultimately, the results of the current study contribute to a growing literature base examining PI, ATL, and student achievement. Although results did not indicate that ATL was a substantial mediator of the effects of PI on academic achievement, observed relationships between PI and academic achievement and between ATL and academic achievement underscore the continued importance of these variables. Further examination of these relationships and exploration of potential mediators of the effects of PI on academic achievement is needed to better support students in schools and their families. References Anthony, C. J., DiPerna, J. C., & Amato, P. R. (2014). Divorce, approaches to learning, and children’s academic achievement: A longitudinal analysis of mediated and moderated effects. Journal of School Psychology, 52, 249 –261. http://dx.doi.org/10.1016/j.jsp.2014.03.003 Ardoin, S. P., & January, S. A. (2018). Academic assessment and intervention. In S. L. Grapin & J. H. Kranzler (Eds.), School psychology: Professional issues and practices (pp. 131–147). New York, NY: Springer. Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. http://dx.doi.org/10.1037/0022-3514.51.6.1173 Bempechat, J., & Shernoff, D. J. (2012). Parental influences on achievement motivation and student engagement. In S. L. Christenson, A. L. Reschly, & C. Wylie (Eds.), Handbook of research on student engagement (pp. 315–342). New York, NY: Springer Science⫹Business Media. http://dx.doi.org/10.1007/978-1-4614-2018-7_15 Bronfenbrenner, U., & Ceci, S. J. (1994). Nature-nurture reconceptualized in developmental perspective: A bioecological model. Psychological Review, 101, 568 –586. http://dx.doi.org/10.1037/0033-295X.101.4.568 Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9, 233–255. http://dx.doi.org/10.1207/S15328007SEM0902_5 Claessens, A., Duncan, G. J., & Engel, M. (2009). Kindergarten skills and fifth grade achievement: Evidence from the ECLS-K. Economics of Education Review, 28, 415– 427. http://dx.doi.org/10.1016/j.econedurev .2008.09.003 Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Mahwah, NJ: Erlbaum. Cole, D. A., & Maxwell, S. E. (2003). Testing mediational models with longitudinal data: Questions and tips in the use of structural equation modeling. Journal of Abnormal Psychology, 112, 558 –577. http://dx.doi .org/10.1037/0021-843X.112.4.558 Cole, D. A., & Preacher, K. J. (2014). Manifest variable path analysis: Potentially serious and misleading consequences due to uncorrected measurement error. Psychological Methods, 19, 300 –315. http://dx.doi .org/10.1037/a0033805 DiPerna, J. C., & Elliott, S. N. (2002). Promoting academic enablers to improve student achievement: An introduction to the mini-series. School Psychology Review, 31, 293–297. DiPerna, J. C., Volpe, R. J., & Elliott, S. N. (2002). A model of academic enablers and elementary reading/language arts achievement. School Psychology Review, 31, 298 –312. DiPerna, J. C., Volpe, R. J., & Elliott, S. N. (2005). A model of academic enablers and mathematics achievement in the elementary grades. Journal of School Psychology, 43, 379 –392. http://dx.doi.org/10.1016/j.jsp .2005.09.002 Duncan, G. J., Dowsett, C. J., Claessens, A., Magnuson, K., Huston, A. C., Klebanov, P., . . . Japel, C. (2007). School readiness and later achievement. Developmental Psychology, 43, 1428 –1446. http://dx.doi.org/10 .1037/0012-1649.43.6.1428 Englund, M. M., Luckner, A. E., Whaley, G. J. L., & Egeland, B. (2004). Children’s achievement in early elementary school: Longitudinal effects of parental involvement, expectations, and quality of assistance. Journal of Educational Psychology, 96, 723–730. http://dx.doi.org/10.1037/ 0022-0663.96.4.723 Epstein, J. L. (1995). School/family/community partnerships: Caring for the children we share. Phi Delta Kappan, 76, 701–712. Fan, W., & Williams, C. (2010). The effects of parental involvement on students’ academic self-efficacy, engagement and intrinsic motivation. Educational Psychology, 30, 53–74. http://dx.doi.org/10.1080/ 01443410903353302 Fan, X., & Chen, M. (2001). Parental involvement and students’ academic achievement: A meta-analysis. Educational Psychology Review, 13, 1–22. http://dx.doi.org/10.1023/A:1009048817385 Fantuzzo, J., McWayne, C., Perry, M. A., & Childs, S. (2004). Multiple dimensions of family involvement and their relations to behavioral and learning competencies for urban, low-income children. School Psychology Review, 33, 467– 480. Fantuzzo, J. W., Tighe, E., & Perry, M. (1999). Relationships between family involvement in Head Start and children’s interactive peer play. NHSA Dialog: A Research-to-Practice Journal for the Early Intervention Field, 3, 60 – 67. http://dx.doi.org/10.1207/s19309325nhsa0301_6 Farrington, C. A., Roderick, M., Allensworth, E., Nagaoka, J., Keyes, T. S., Johnson, D. W., & Beechum, N. O. (2012). Teaching adolescents to become learners. The role of noncognitive factors in shaping school performance: A critical literature review. Chicago, IL: University of Chicago Consortium on Chicago School Research. Galindo, C., & Sheldon, S. B. (2012). School and home connections and children’s kindergarten achievement gains: The mediating role of family involvement. Early Childhood Research Quarterly, 27, 90 –103. http:// dx.doi.org/10.1016/j.ecresq.2011.05.004 Garbacz, S. A., Herman, K. C., Thompson, A. M., & Reinke, W. M. (2017). Family engagement in education and intervention: Implementa- This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. PARENT INVOLVEMENT, ATL, AND ACHIEVEMENT tion and evaluation to maximize family, school, and student outcomes. Journal of School Psychology, 62, 1–10. http://dx.doi.org/10.1016/j.jsp .2017.04.002 Garbacz, S. A., McDowall, P. S., Schaughency, E., Sheridan, S. M., & Welch, G. W. (2015). A multidimensional examination of parent involvement across child and parent characteristics. The Elementary School Journal, 115, 384 – 406. http://dx.doi.org/10.1086/680325 Grolnick, W. S., & Slowiaczek, M. L. (1994). Parents’ involvement in children’s schooling: A multidimensional conceptualization and motivational model. Child Development, 65, 237–252. http://dx.doi.org/10 .2307/1131378 Hoover-Dempsey, K. V., Walker, J. M. T., Sandler, H. M., Whetsel, D., Green, C., Wilkins, A. S., & Closson, K. (2005). Why do parents become involved? Research findings and implications. The Elementary School Journal, 106, 105–130. http://dx.doi.org/10.1086/499194 Hu, L., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55. http://dx.doi.org/10.1080/10705 519909540118 Izzo, C. V., Weissberg, R. P., Kasprow, W. J., & Fendrich, M. (1999). A longitudinal assessment of teacher perceptions of parent involvement in children’s education and school performance. American Journal of Community Psychology, 27, 817– 839. http://dx.doi.org/10.1023/A:10222 62625984 Jeynes, W. H. (2005). A meta-analysis of the relation of parent involvement to urban elementary school student academic achievement. Urban Education, 40, 237–269. http://dx.doi.org/10.1177/0042085905274540 Jeynes, W. (2012). A meta-analysis of the efficacy of different types of parental involvement programs for urban students. Urban Education, 47, 706 –742. http://dx.doi.org/10.1177/0042085912445643 Li-Grining, C. P., Votruba-Drzal, E., Maldonado-Carreño, C., & Haas, K. (2010). Children’s early approaches to learning and academic trajectories through fifth grade. Developmental Psychology, 46, 1062–1077. http://dx.doi.org/10.1037/a0020066 Little, T. D. (2013). Longitudinal structural equation modeling. New York, NY: Guilford Press. Mahatmya, D., Lohman, B. J., Matjasko, J. L., & Farb, A. F. (2012). Engagement across developmental periods. In S. L. Christenson, A. L. Reschly, & C. Wylie (Eds.), Handbook of research on student engagement (pp. 45– 63). New York, NY: Springer. http://dx.doi.org/10.1007/ 978-1-4614-2018-7_3 Manz, P. H., Fantuzzo, J. W., & Power, T. J. (2004). Multidimensional assessment of family involvement among urban elementary students. Journal of School Psychology, 42, 461– 475. http://dx.doi.org/10.1016/ j.jsp.2004.08.002 Maxwell, S. E., Cole, D. A., & Mitchell, M. A. (2011). Bias in crosssectional analyses of longitudinal mediation: Partial and complete mediation under an autoregressive model. Multivariate Behavioral Research, 46, 816 – 841. http://dx.doi.org/10.1080/00273171.2011.606716 McIntyre, L. L., Eckert, T. L., Fiese, B. H., DiGennaro, F. D., & Wildenger, L. K. (2007). Transition to kindergarten: Family experiences and involvement. Early Childhood Education Journal, 35, 83– 88. http:// dx.doi.org/10.1007/s10643-007-0175-6 McWayne, C., Hampton, V., Fantuzzo, J., Cohen, H. L., & Sekino, Y. (2004). A multivariate examination of parent involvement and the social and academic competencies of urban kindergarten children. Psychology in the Schools, 41, 363–377. http://dx.doi.org/10.1002/pits.10163 Muthén, L. K., & Muthén, B. O. (1998 –2017). Mplus user’s guide (8th ed.). Los Angeles, CA: Author. Najarian, M., Pollack, J. M., & Sorongon, A. G. (2009). Early childhood longitudinal study, kindergarten class of 1998 –99 (ECLS-K): Psycho- 385 metric report for the eighth grade (NCES 2009 – 002). Washington, DC: National Center for Education Statistics, Institute of Education Sciences, U.S. Department of Education. Preacher, K. J., & Selig, J. P. (2012). Advantages of Monte Carlo confidence intervals for indirect effects. Communication Methods and Measures, 6, 77–98. http://dx.doi.org/10.1080/19312458.2012.679848 Prenoveau, J. M. (2016). Specifying and interpreting latent state–trait models with autoregression: An illustration. Structural Equation Modeling, 23, 731–749. http://dx.doi.org/10.1080/10705511.2016.1186550 R Core Team. (2016). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from https://www.R-project.org Roorda, D. L., Koomen, H. M., Spilt, J. L., & Oort, F. J. (2011). The influence of affective teacher–student relationships on students’ school engagement and achievement: A meta-analytic approach. Review of Educational Research, 81, 493–529. http://dx.doi.org/10.3102/003 4654311421793 Selig, J. P., & Preacher, K. J. (2009). Mediation models for longitudinal data in developmental research. Research in Human Development, 6, 144 –164. http://dx.doi.org/10.1080/15427600902911247 semTools Contributors. (2016). semTools: Useful tools for structural equation modeling. R package version 0.4 –14. Retrieved from https://CRAN .R-project.org/package⫽semTools Thompson, B. C., Mazer, J. P., & Flood Grady, E. (2015). The changing nature of parent–teacher communication: Mode selection in the smartphone era. Communication Education, 64, 187–207. http://dx.doi.org/10 .1080/03634523.2015.1014382 Tourangeau, K., Lê, T., & Nord, C. (2005). Early Childhood Longitudinal Study, Kindergarten Class of 1998 –99 (ECLS-K), Fifth grade methodology report (NCES 2006 – 037). Washington, DC: National Center for Education Statistics. Tourangeau, K., Nord, C., Lê, T., Sorongon, A. G., & Najarian, M. (2009). Combined user’s manual for the ECLS-K eighth grade and K– 8 full sample data files and electronic codebooks (NCES 2009 – 004). Washington, DC: U.S. Department of Education, National Center for Education Statistics. Urdan, T., & Schoenfelder, E. (2006). Classroom effects on student motivation: Goal structures, social relationships, and competence beliefs. Journal of School Psychology, 44, 331–349. http://dx.doi.org/10.1016/j .jsp.2006.04.003 Wang, M. C., Haertel, G. D., & Walberg, H. J. (1993). Toward a knowledge base for school learning. Review of Educational Research, 63, 249 –294. http://dx.doi.org/10.3102/00346543063003249 White, I. R., Royston, P., & Wood, A. M. (2011). Multiple imputation using chained equations: Issues and guidance for practice. Statistics in Medicine, 30, 377–399. http://dx.doi.org/10.1002/sim.4067 Xu, M., Kushner Benson, S. N., Mudrey-Camino, R., & Steiner, R. P. (2010). The relationship between parental involvement, self-regulated learning, and reading achievement of fifth grades: A path analysis using the ECLS-K database. Social Psychology of Education, 13, 237–269. http://dx.doi.org/10.1007/s11218-009-9104-4 Youn, M. J., Leon, J., & Lee, K. J. (2012). The influence of maternal employment on children’s learning growth and the role of parental involvement. Early Child Development and Care, 182, 1227–1246. http://dx.doi.org/10.1080/03004430.2011.604944 Received April 5, 2018 Revision received July 23, 2018 Accepted July 30, 2018 䡲