Chapter 7 Experimental Design I— Independent Variables Copyright © 2016 Wolters Kluwer • All Rights Reserved The Research Process Evidence-based practice If found The Body of Knowledge • Anecdotal observations • Scientific literature base Search for answers If not found Design and development of a study • Funding • Human and animal use approval • Pilot studies • Preliminary data If accepted Publication of manuscript Peer review If rejected Manuscript preparation Conduct of the study Statistical analysis of data Collection of the data Laboratory analysis of data Copyright © 2016 Wolters Kluwer • All Rights Reserved 2 Key Terms Alternate/research hypothesis Inclusion criteria Between-group/-subject independent Independent variable variable Internal validity Bias Multivariate Blind Non-experimental research Confounding or intervening variable Null hypothesis Constant Placebo effect Control group Power Counterbalancing Quasi-experimental research Dependent variable Research design Double blind Treatment/intervention Exclusion criteria Univariate Experimental research Variable External validity Within-group/-subject independent Hawthorne effect variable Copyright © 2016 Wolters Kluwer • All Rights Reserved 3 Planning the Research Design • Research design is the process by which investigators determine how to answer their research question(s) • Flaws in research design typically cannot be overcome by editing or statistical analysis Copyright © 2016 Wolters Kluwer • All Rights Reserved 4 Identifying Variables • A variable is some characteristic or factor that can have different values and is either subject to change or can be manipulated as an intervention • Variables may be independent, dependent, constant, or confounding Copyright © 2016 Wolters Kluwer • All Rights Reserved 5 Identifying Variables Independent Variable Dependent Variable Intentionally controlled or manipulated Outcome for which researcher hopes to elicit an effect Treatment or intervention “What is measured” May be multiple levels May be group classification Unknown factor Known factor Copyright © 2016 Wolters Kluwer • All Rights Reserved 6 Independent Variables: Levels • Examining the effects of different doses of a drug or supplement is an example of multiple levels of a single independent variable • How different doses affect male and female mice is an example of multiple independent variables Fig. 7-1 Copyright © 2016 Wolters Kluwer • All Rights Reserved 7 Independent Variables: Types • Between-group (or between-subjects) independent variable: different group of subjects for each level of the variable • Within-group (or withinsubject) independent variable: each subject is tested at each level of the independent variable Fig. 7-1 Copyright © 2016 Wolters Kluwer • All Rights Reserved 8 Identifying Variables Between-Group/-Subject Within-Group/-Subject Well suited to animal studies Each subject serves as his/her own control More demanding of the participants Random group assignment can reduce within-group variability; matching might be preferred Powerful statistical design Reduces sample size and need for randomization Requires longer time period, including accounting for wash-out Copyright © 2016 Wolters Kluwer • All Rights Reserved 9 Overview of Research Design Dimensions Design Primary Use Randomized? Degree of Retrospective or Emphasis on Control Prospective? Validity NonExperimental Description, examine relationships No Low Either QuasiExperimental Causal inferences No Low/mod erate Either External True Experimental Causal inferences Yes High Prospective Internal Table 7-4 Copyright © 2016 Wolters Kluwer • All Rights Reserved 10 Experimental Design: Control Group • Control group: measured at the same time points as the treatment group(s) but receives no treatment • Placebo: dummy treatment that does not affect the dependent variable(s) • Blinding: keeping participants (single blind) and ideally both participants and study personnel (double blind) naïve to the study treatment to limit bias Copyright © 2016 Wolters Kluwer • All Rights Reserved 11 Experimental Design: Control Group • Placebo effect: subjects receiving the placebo may experience a benefit even though they aren’t receiving any treatment • Hawthorne effect: subjects may perform better due to being observed Copyright © 2016 Wolters Kluwer • All Rights Reserved 12 Experimental Design: Selecting Subjects • Identify the study population, and obtain a representative sample • Weigh the ability to improve retention (convenience sample) against having a more representative sample • The degree to which the sample represents the population of interest affects the power of the study • Inclusion criteria are the stated subject characteristics • Exclusion criteria restrict subject participation Copyright © 2016 Wolters Kluwer • All Rights Reserved 13 Experimental Design: Controlling Variability & Bias • Using randomization or matching to assign subjects to groups reduces other factors affecting the dependent variable • Controlling for confounding or intervening variables reduces threats to internal validity • Minimizing signal-to-noise ratio better enables the impact of the independent variable on the dependent variable to be observed • Systematic errors occur when measurement error is in one direction • Random errors may occur in any direction and typically have a net zero effect Copyright © 2016 Wolters Kluwer • All Rights Reserved 14 Summary • Research design requires balance, weighing the pros and cons of a number of experimental choices • There is a trade-off between controlling variables and realworld applicability • Planning is key for avoiding confounding factors Copyright © 2016 Wolters Kluwer • All Rights Reserved 15