4
Experimental Research Methods (Fach) / L8 (Lektion)
After this lecture and studying the materials, students are able to:
• Explain the differences between one-way ANOVA, two-way ANOVA, and ANCOVA
• Interpret the SPSS output of an ANCOVA using a quasi-experimental design
• Explain how ANCOVA can eliminate bias
Diese Lektion wurde von tklmn erstellt.
Diese Lektion ist leider nicht zum lernen freigegeben.
- ANCOVA = ANalysis of COVAriance • Do population means differ from each other controlled for covariate X?• Does factor A have an effect, controlled for covariate X?• Statistical control rather than experimental control• ANCOVA is interpreted in the same way as one-way ANOVA: the main effect of factor A but thencorrected for covariance X
- ANCOVA: Idea • You are interested in the effect of one factor, but you want control for the covariate as well• For example, the effect of type of education on learning performance controlled for IQ• Covariate = continuous variable that is related to the dependent variable that cannot beexperimentally controlled for• When levels of a factor, on average, differ on the covariate, there will also be a mean difference onthe DV (=bias) that has nothing to do with the factor• Using an ANCOVA eliminates bias, and we can answer the questions:• Would the means of the dependent variable be different across the levels of the factor, if the means of the covariate wereequal?• Is there an effect of the factor, controlled for the covariate?
- Reduction of error variance • Whether error variance (MSw) is reduced depends on the relation between the covariate X and Y• This is also the case in regression analysis• Reduction of error variance is the largest when thecorrelation is very large (positive) or small (negative)• Reduction of error variance yields larger power, because f goes up if msw goes own
- Interpretation quasi-experiments • There is a danger in the interpretation of statistical analysis (so also of ANCOVA) of existing groups• Internal validity may be low• For example, a quasi-experiment may show that energy drink has an effect on exam grade but notwhen taking the covariate hours preparation into account• And there might be other confounding variables that explain the difference (“bias”)• Hence, ANCOVA is OK, but keep threats to the internal validity in mind
Anzeige