2 Way ANOVA, Post Hoc Tests

The data set for this exercise is: 2WayAnovaR.sav

• Use the variable label information to figure out what each variable is, and its possible range.

• Use Frequencies to identify any data that seems to have been entered inaccurately. If so, decide what to do about that.

1. Are each of the three variables of interest continuous or categorical?
2. Of attitude and peer pressure, which is the stronger predictor of intention?
3. What is the median for attitudes and the median for peer pressure?

• Using “Transform> Recode > Recode into different variables”, conduct median splits on both variables.
  • Don't forget to include the actual median scores with either the "high" or the "low" group.
    
    
• Conduct a 2 way ANOVA.

• Ask for the "estimated marginal means" for the main effects and interaction.
• Ask for estimates of effect size (partial eta-squared).

4a.  Look at the post hoc menu options. Should you check anything here? Why or why not?

• Under the plots menu, ask SPSS to generate at least one plot of the interaction, deciding for yourself which variable should go on the horizontal and which should be represented by separate lines.

4b.  Is each main effect significant? Is the interaction significant?

4c.  Write a paragraph reporting on the results of this analysis, use proper APA style (don’t forget to include effect size and, for significant findings, the means).

• If the interaction is significant, conduct post hoc tests by using the following method:
  • Use “Transform>Compute” and the “If” option to create a new variable labeling the experimental condition for each participant.  That is, this variable will have 4 possible values, each representing one cell in the factorial design.
  • Use that variable as the independent variable to conduct a one-way ANOVA on intention.
  • Ask for Tukey post hoc tests.
  • Interpret the results of each post hoc test.

5. Add a sentence or two to your paragraph from #4c  indicating that:  “A ______ post-hoc test revealed that ……” (describe outcome, e.g,. are all four means different from one another?).

ANCOVA

The data set for this exercise is: ANCOVA.sav

This experiment measured the extent to which the female participants base their self-esteem on the reflected appraisals of other people (10 item scale, each item labeled r1, r2, etc.), and measured participants’ trait self-esteem (rosen). Next, either positive or negative feedback was given to the participant. Last, researchers measured temporary state self-esteem related to their feelings about how they looked (appse).

• Use the variable label information to figure out what each variable is, and its possible range.

• Use Frequencies to identify any data that seems to have been entered inaccurately. If so, decide what to do about that.

• Create the reflected appraisals scale from the individual items.
  • Recode reflected appraisals items as needed.

6a. Write down which item numbers you recoded.

• Compute the internal consistency of the reflected appraisals scale.

6b. Write a sentence reporting whether the reliability of the scale is adequate or not (report the alpha).

• Create the scale using Transform>Compute and the Function for “MEAN”

    7.  Test the hypothesis that negative feedback leads to lower state self-esteem (appse) and write an APA style paragraph reporting your conclusions.

• The actual hypothesis of the study is that basing worth on reflected appraisals will moderate the relationship between feedback and state self-esteem (Hint: think about what a moderator variable is in the vocabularly of an ANOVA analysis).
  
  
  • Do the necessary transformations to the data to test that hypothesis with a 2 x 2 ANOVA. (Hint: reflected appraisals is currently a continuous variable.)

    8a.  In the 2-way analysis, does feedback have any impact on state self-esteem? 

    8b.  Why were you unable to see the influence of feedback on state self-esteem in the analysis from #7? (Hint: comment on the informative nature of interactions.)

• Rosen (trait self-esteem) is probably an extraneous variable related to the dependent variable.

9a.  Confirm or disconfirm this with a bivariate correlation (include the correlation in APA style with your response).

9b.  Assuming rosen is correlated with appse, why should we probably do an ANCOVA rather than making rosen the third variable in a 2 x 2 x 2 ANOVA? (Hint: this is a conceptual, not statistical question)

• Conduct the 2 x 2 ANCOVA.

10.  Write a paragraph explaining the analysis and results. The following should be the first sentence of your paragraph: “We conducted a 2 (hi/low reflected appraisals) x 2 (positive/negative feedback) ANCOVA with trait self-esteem as the covariate.” Report on whether trait self-esteem “was a significant covariate” (or not) first.  Then report on the main effects and then interactions including all the proper information in APA style.