Psyc611

Assignments

1. Assignments parallel the lectures. The graduate assistant will assign deadlines to which you should adhere, and will adjust grades for lateness as appropriate.  Be advised that the graduate assistant will be required to submit grades for assignments 1-7 by the time of the midterm exam, and for assignments 8-12 by the time of the endterm examination

2. A maximum of 2 points for each question will be assigned (i.e. 1a, 1b, 1c counts as a single question).  The point scheme is 0 = incomplete/missing, 1 = late/minor errors, 2 perfect.  The assignment grade comprises 10% of the final grade, and will derive from the sum of these points.  The maximum points are 64.

3. Be advised that these assignments are not intended as collaborative exercises.  Your work should be your own.  If you hit a snag, you are encouraged to solicit the assistance of the graduate assistant.

4. Unless otherwise noted, use the main data set.

Assignment #1 Data Entry and Variable Naming

1.              Enter the main data set (posted on the web) and run descriptives to see if data look "clean." That is,are any data outside the minimum or maximum for the scale? (Make sure to include a label for gender, so that you remember -1 is male)

Assignment #2 Descriptive Statistics and Data Transformations

1a.            Examine the skew of RT.

1b.            Transform RT with log10 and with the reciprocal, then reassess skew (Don't lose the original data!!)

1c.            Assume that you recheck your raw data, and find that 65.50 should be 6.50 (replace and save correct score in the data file).  Reassess skew.

2a.            Reverse score eval_5.  Save it under the same name

2b.           Assess alpha for the 5 item evaluation scale

2c.            Compute a new variable, evalscal, from the 5 items

2d.           Imagine that in version 1 of the survey, frstr is a rating of Mary and sectry is a rating of  Mark.  In version -1, the opposite is true (i.e., target gender was counterbalanced). Create new variables called Mary and Mark.

Assignment #3 Quick-and-Dirty ANOVAs

1a.            Compute a 1-way between groups ANOVA using manip as IV and rt as DV Remember: You corrected the rt data file in Assignment #2

1b.            Show Levene's test for homogeneity of variance

1c.            What other evidence might you use for homogeneity?

2a.            Compute a 2-way between groups ANOVA using manip and gender as IVs and eval_1 as the DV

2b.           Show the 6 cell means and standard deviations in an APA style table

3a.            Show the 2-way mixed ANOVA (S/AxB) using target gender (Mary and Mark) within and subject gender between.

3b.           Graph the 2-way interaction from 3a

Assignment #4 Repeated Measures ANOVA and Epsilon

1a.            Compute 1-way within subjects ANOVA using eval_1, eval_2, eval_3,  eval_4, and (reverse-scored) eval_5.

1b.            Find the variance-covariance matrix for the 5 levels of eval (from 1a).

1c.            Compute the Geisser-Greenhouse estimate for epsilon by hand and also compute the minimum by hand.  Compare to the SPSS output from 1a. Compute the Huynh-Feldt estimate for epsilon by hand and compare to the SPSS output from 1a.

2.              Using 2-way mixed ANOVA (S/AxB) show how you would test whether the counterbalancing of Mary and Mark produced a carry-over effect

Assignment #5 Power, Effect Sizes, Agreement Indices

1.              Using gender as the IV and rt as the DV:

1a. find power using SPSS

1b. find power using psylib

1c. find power using f and the corresponding power table

1d. using Dunlap's shortcut, what n would give 80-90% power?

2.              Using gender as the IV and rt as the DV (as above)

2a. show h2 both by hand and in SPSS

2b. show w2 by hand, using the SPSS ANOVA output

2c. show the 95% confidence interval around the difference, by hand and on SPSS

3.              jcateg and kcateg are two judges' categorization of participants' verbal responses into 1 of 3 mutually exclusive and exhaustive categories.  Show k by hand and using SPSS

4.              now imagine that jcateg and kcateg are interchangeable judges' ratings on a 1-3 scale.  Show the intraclass correlation coefficient both in SPSS and by hand, and its test of significance

Assignment #6 Tests Subsequent to ANOVA-- Main Effects

1a.            Compute a 1-way between groups ANOVA using manip as IV and rt as DV (from assignment #3)

1b.            Show Tukey HSD post-tests

2a.            Show the orthogonal contrasts of M1 vs M2 as well as M3 vs (M1+M2)/2  Use the one-way command so that you can find SSs for the contrasts. Compute the SS for the first contrast by hand, and compare to the SPSS output

2b.           Show the linear and quadratic trends.

2c.            What would the polynomial coefficients for the linear and quadratic trends be if you had 5 groups?

3.              Compute a 1-way repeated measures ANOVA using eval_2, eval_3, and eval_4 as the levels of your within subjects variable.  Show the linear and quadratic trends.

Assignment #7 Tests Subsequent to ANOVA -- Interactions

1.               Compute a 2x3 between groups ANOVA using gender and manip as the IVs and rt as the DV.  Does SPSS allow you to conduct range tests comparing the 6 means of the interaction?  "Trick" the program into doing the Tukey HSD tests for you.

2.              Show the simple effects tests of manipulation for each level of gender (DV is rt).

3.              Imagine that the quadratic trend of eval1, eval3, and eval4 is expected to vary across gender.  Show the test in SPSS.

4.              Use the Tukey HSD procedure to compare the 6 means from the S/AxB ANOVA gender-by-evaluation item.  Make sure to use the Welch-Satterthwaite correction. (Use psylib range or compute by hand, rather than using SPSS)

Assignment #8 More Advanced ANOVA designs

1a.            using manip as the IV and attentio as the DV, prove to yourself that ANOVA is the same in one-way designs for fixed and random effects

1b.            using manip and granny as the IVs and prejudic as the DV, conduct the 2-way ANOVA using both as random effects IVs and then both as fixed effect IVs . Compare the output

2b.           use score2 as the DV, and conduct an ANOVA in which schools are nested in systems

Assignment #9 Simple Correlation and Regression

1a.            Regress evalscal (your 5 item scale) on prejudic, saving residuals and the influence statistics.  Show the scatterplot. Comment

1b.            Show the regression equation.  Any thoughts about the fact that R is positive and B is negative?   How did SPSS find the standard error of the estimate?

Assignment #10 Multiple Regression

1a.            You predict that intellig and attentio predict evalscale. Look for signs of multicollinearity, outliers, and suppressor variables.  Plot the  residuals

1b.            Assuming things look okay in 1b, examine the standard multiple regression output and interpret the results

2.              Add Mary to the equation, then correct for shrinkage

3.              Now use hierarchical multiple regression, regressing evalscale on intellig in the first step and adding attentio in the second step.  Interpret your findings.

Assignment #11 Multiple Regression

1a.            Download the mediator/moderator data to test the hypothesis that the interaction of pred1 and moderato predicts dvmod. Start with the usual snooping for problems. Use these data for the entire problem #11

1b.            test the hypothesis (make sure to center continuous predictors)

1c.            if the interaction is significant (or marginal), graph the interaction.

2a.            Test whether mediator mediates the effect of pred2 on dvmed.

2b.           If mediation is apparent, follow-up with the BK modification of Sobel's test

2c.            Represent the results in a figure

3.              Imagine that you wanted to examine pred3 as a moderator of the pred1-->dvmod relation.  Imagine that the entered codes represent three ethnic groups.  How would you include it as a predictor?

Assignment #12 Repeated Measures Multiple Regression

1.             Treat dvmod and dvmed as 2 levels of a within subjects variable "item," and run the repeated measures ANOVA.  Now use repeated measures multiple regression to test the same hypothesis (with rounding error, your F-ratio should be similar and df the same).

2.             Center pred1, and create the pred*iteminteraction term.  Conduct RMMR.

3.             Run a simple regression using centered pred 1 as the predictor and the average of dvmod & dvmed as the criterion. Compare to its corresponding answer in #2 (with rounding error, your F-ratio should be similar and the df the same).

4.              Generate the equations for the cpred*item interaction, and produce the values that you would use to graph it.

*******Revised #12

1.              Treat dvmod and dvmed as 2 levels of a within subjects variable "item," and run the repeated measures ANOVA.  That's just like doing an SxA.  If it helps to think of dvmod/dvmed as time1/time2, go ahead and relabel them as such

Now use repeated measures multiple regression to test the same hypothesis, i.e., that there is a difference between levels.  You have the F values already, but remember you want to report R2 for the model.

2.             Center pred1, and create the pred*iteminteraction term.  Conduct RMMR testing the effects of "item", "centered pred1", and the interaction term.  You can do this procedure in GLM, just make sure to enter the continuous variable as a covariate and tell the program to create the interaction term. You are looking for the F for the change in R2 at each step, the change in R2 for each step, and the final R2 for the equation.

3.             Run a simple regression using centered pred 1 as the predictor and the average of dvmod & dvmed as the criterion. Compare to its corresponding answer in #2 (with rounding error, your F-ratio should be similar and the df the same).

4.              Generate the equations for the cpred*item interaction, and produce the values that you would use to graph it.