國立高雄師範大學九十九學年度博士班招生考試試題 系所別:教育學系 科 (以鉛筆作答者不予計分) 目:英文教育名著(D 類)(第一頁,共五頁) I. Summarize and comment on the following paragraphs in Chinese. (25%) Although data transformations are recommended as a remedy for outliers and for failures of normality, linearity, and homoscedasticity, they are not universally recommended. The reason is that an analysis is interpreted from the variables that are in it, and transformed variables are sometimes harder to interpret. For instance, although IQ scores are widely understood and meaningfully interpreted, the logarithm of IQ scores may be harder to explain. Whether transformation increases difficulty of interpretation often depends on the scale in which the variable is measured. If the scale is meaningful or widely used, transformation often hinders interpretation, but if the scale is somewhat arbitrary anyway (as is often the case), transformation does not notably increase the difficulty of interpretation. With ungrouped data, it is probably best to transform variables to normality unless interpretation is not feasible with the transformed scores. With grouped data, the assumption of normality is evaluated with respect to the sampling distribution of means (not the distribution of scores) and the Central Limit Theorem predicts normality with decently sized samples. However, transformations may improve the analysis and may have the further advantage of reducing the impact of outliers. Our recommendation, then, is to consider transformation of variables in all situations unless there is some reason not to. If you decide to transform, it is important to check that the variable is normally or near-normally distributed after transformation. Often you need to try first one transformation and then another until you find the transformation that produces the skewness and kurtosis values nearest zero, the prettiest picture, and/or the fewest outliers. With almost every data set in which we have used transformations, the results of analysis have been substantially improved. This is particularly true when some variables are skewed and others are not, or variables are skewed very differently prior to transformation. However, if all the variables are skewed to about the same moderate extent, improvements of analysis with transformation are often marginal. With grouped data, the test of mean differences after transformation is a test of differences between medians in the original data. After a distribution is normalized by transformation, the mean is equal to the median. The transformation affects the mean but not the median because the median depends only on rank order of cases. Therefore, (背面有題 續翻背面) 第 1 頁,共 5 頁 系所別:教育學系 科 (以鉛筆作答者不予計分) 目:英文教育名著(D 類)(第二頁,共五頁) conclusions about means of transformed distributions apply to medians of untransformed distributions. Transformation is undertaken because the distribution is skewed and the mean is not a good indicator of the central tendency of the scores in the distribution. For skewed distributions, the median is often a more appropriate measure of central tendency than the mean, anyway, so interpretation of differences in medians is appropriate. II. Summarize and comment on the following paragraphs in Chinese. (25%) Underlying some multivariate procedures and most statistical tests of their outcomes is the assumption of multivariate normality. Multivariate normality is the assumption that each variable and all linear combinations of the variables are normally distributed. When the assumption is met, the residuals of analysis are also normally distributed and independent. The assumption of multivariate normality is not readily tested because it is impractical to test an infinite number of linear combinations of variables for normality. Those tests that are available are overly sensitive. The assumption of multivariate normality is made as part of derivation of many significance tests. Although it is tempting to conclude that most inferential statistics are robust to violations of the assumption, that conclusion may not be warranted. Bradley (1982) reports that statistical inference becomes less and less robust as distributions depart from normality, rapidly so under many conditions. And even when the statistics are used purely descriptively, normality, linearity, and homoscedasticity of variables enhance the analysis. The safest strategy, then, is to use transformations of variables to improve their normality unless there is some compelling reason not to. The assumption of multivariate normality applies differently to different multivariate statistics. For analyses when subjects are not grouped, the assumption applies to the distributions of the variables themselves or to the residuals of the analyses; for analyses when subjects are grouped, the assumption applies to the sampling distributions of means of variables. If there is multivariate normality in ungrouped data, each variable is itself normally distributed and the relationships between pairs of variables, if present, are linear and homoscedastic (i.e., the variance of one variable is the same at all values of the other variable). The assumption of multivariate normality can be partially checked by examining the normality, linearity, and homoscedasticity of individual variables or through examination of residuals in analyses involving prediction. The assumption is certainly violated, at least to some extent, if the individual variables (or the residuals) are not normally distributed or do not have pairwise linearity and homoscedasticity. 第 2 頁,共 5 頁 系所別:教育學系 科 (以鉛筆作答者不予計分) 目:英文教育名著(D 類)(第三頁,共五頁) For grouped data, it is the sampling distributions of the means of variables that are to be normally distributed. The Central Limit Theorem reassures us that, with sufficiently large sample sizes, sampling distributions of means are normally distributed regardless of the distributions of variables. For example, if there are at least 20 degrees of freedom for error in a univariate ANOVA, the F test is said to be robust to violations of normality of variables (provided that there are no outliers). III. Summarize and comment on the following paragraphs in Chinese. (25%) Development Is Contextual All development occurs within a context, or setting. Contexts include families, schools, peer groups, churches, cities, neighborhoods, university laboratories, countries, and so on. Each of these settings is influenced by historical, economic, social, and cultural factors (Shiraev & Levy, 2007; Yang, 2005). Contexts, like individuals, change. Thus, individuals are changing beings in a changing world. As a result of these changes, contexts exert three types of influences (Baltes, 2000, 2003): (1) normative age-graded influences, (2) normative history-graded influences, and (3) nonnormative or highly individualized life events. Each of these types can have a biological or environmental impact on development. Normative age-graded influences are similar for individuals in a particular age group. These influences include biological processes such as puberty and menopause. They also include sociocultural, environmental processes such as beginning formal education (usually at about age 6 in most cultures) and retirement (which takes place in the fifties and sixties in most cultures). Normative history-graded influences are common to people of a particular generation because of historical circumstances. For example, in their youth, American baby boomers shared the experience of the Cuban missile crisis, the assassination of John F. Kennedy, and the Beatles invasion. Other examples of normative history-graded influences include economic, political, and social upheavals such as the Great Depression in the 1930s, World War II in the 1940s, the civil rights and women’s rights movements of the 1960s and 1970s, the terrorist attacks of 9/11/2001, as well as the integration of computers and cell phones into everyday life during the 1990s (Modell & Elder, 2002; Schaie, 2007). Long-term changes in the genetic and cultural makeup of a population (due to immigration or changes in fertility rates) are also part of normative historical change. (背面有題 續翻背面) 第 3 頁,共 5 頁 系所別:教育學系 科 (以鉛筆作答者不予計分) 目:英文教育名著(D 類)(第四頁,共五頁) Nonnormative life events are unusual occurrences that have a major impact on the individual’s life. These events do not happen to all people, and when they do occur they can influence people in different ways. Examples include the death of a parent when a child is young, pregnancy in early adolescence, a fire that destroys a home, winning the lottery, or getting an unexpected career opportunity with special privileges. In the life of Alice Walker described at the opening of the chapter, being shot in the eye was a nonnormative life event. IV. Translate followings into Chinese and write your comments (either English or Chinese). (25%) Evolutionary Developmental Psychology Recently, interest has grown in using the concepts of evolutionary psychology to understand human development (Geary, 2006). Here are a few ideas proposed by evolutionary developmental psychologists (Bjorklund & Pellegrini, 2002, pp. 336-340): An extended juvenile period evolved because humans require time to develop a large brain and learn the complexity of human social communities. Humans take longer to become reproductively mature than any other mammal (see figure 3.1). During this juvenile period they develop a large brain and the experiences required for mastering the complexities of human society. “Many aspects of childhood function as preprations for adulthood and were selected over the course of evolution” (p. 337). Play is one possible example. Beginning in the preschool years, boys in all cultures engage in more rough-and-tumble play than girls. Perhaps rough-and-tumble play prepares boys for fighting and hunting as adults. In contrast to boys, girls engage in play that involves more imitation of parents, such as caring for dolls, and less physical dominance. This, according to evolutionary psychologists, is an evolved tendency that prepares females for becoming the primary caregivers for their offspring. Some characteristics of childhood were selected because they are adaptive at specific points in development, not because they prepare children for adulthood. For example, some aspects of play may function, not to prepare us for adulthood, but to help children adapt to their immediate circumstances, perhaps to learn about their current environment. Many evolved psychological mechanisms are domain-specific. That is, the mechanisms apply only to a specific aspect of a person’s makeup (Atkinson & Wheeler, 2004; Rubenstein, 2004). According to evolutionary psychology, information processing is one example. In this view, the mind is not a general-purpose device that can be applied equally to a vast array of problems. Instead, as our ancestors dealt with certain 第 4 頁,共 5 頁 系所別:教育學系 科 (以鉛筆作答者不予計分) 目:英文教育名著(D 類)(第五頁,共五頁) recurring problems, specialized modules evolved that process information related to those problems, such as a module for physical knowledge, a module for mathematical knowledge, and a module for language. Also in this view, “infants enter the world ‘prepared’ to process and learn some information more readily than others, and these preparations serve as the foundation for social and cognitive development” (p. 338). For example, much as goslings in Lorenz’ experiment (described in chapter 2) were “prepared” to follow their mother, human infants are biologically prepared to learn the sounds that are part of human language. Evolved mechanisms are not always adaptive in contemporary society. Some behaviors that were adaptive for our prehistoric ancestors may not serve us well today. For example, the food-scarce environment of our ancestors likely led to humans’ propensity to gorge when food is available and to crave high-caloric foods, a trait that might lead to an epidemic of obesity when food is plentiful. 第 5 頁,共 5 頁