PSY 5510 HA 04 Analyses involving qualitative and quantitative predictors

1. This assignment is based on Jody Damron’s thesis, completed about 10 years ago. Jody gave two personality questionnaires to just under 400 students. All students were instructed to respond honestly to the first questionnaire and to fake in one of four ways on the 2nd. The four different ways of faking defined four conditions of the research.

The data are in Damron Subset 2.

Here is how he described the four: “In the fake instruction condition, participants were asked to either attempt to score as highly as possible on the scale, or to distort their responses in an attempt to gain employment in one of three positions: manager, sales, accountant.”

The variable, avsdfake, for “Average Standardized Difference FAKE score is a measure of how much respondents changed their responses from the 1st to the 2nd administration of the personality questionnaires.

The variable, condit, represents the four conditions: 1=Pure Faking; 2 = Fake to be a manager; 3 = Fake to be a sales person, 4 = Fake to be an accountant.

The variable cog_ab represents Wonderlic Personnal Test scores – cognitive ability.

Your task is to compare mean avsdfake scores across the four conditions, controlling for cog_ab.

That is, you’ll assess

1) the main effect of condit (use whatever comparisons you with),

2) the main effect of cog_ab, and

3) the interaction of condit and cog_ab.

For this problem, leave the interaction in the analysis when testing main effects.

Test the above 3 hypotheses using SPSS GLM and using RCMDR Linear Models procedures.

A. Submit GLM output with the Tests of Between-subject Table, and Parameter estimates table.

B. Submit output of results of analyses using RCMDR.

Each submission should show the same results. That is, it should not matter which program you use, you should reach the exact same conclusion for each test.

2. Validity of Inconsistency. For this problem, use the file, GFP1_Sebren_110813.sav . This file contains real data.

Perform analyses as described below to answer the following questions . . .

The first question is: Does response inconsistency predict performance over and above, i.e., controlling for, cognitive ability and conscientiousness.

The second question is: Does response inconsistency moderate the conscientiousness -> performance relationship? That is, is conscientiousness a better or worse predictor for people who are not inconsistent in their responses than it is for people who respond inconsistently?

In one study (Reddock, Biderman, & Nguyen, 2011 on my web site) we have found that inconsistency is a valid predictor and offers incremental validity over and above both cognitive ability and conscientiousness. We also found that it moderated the conscientiousness -> performance relationship. This homework assignment involves a different set of data and an attempt to replicate that original result.

Consistency

Our measure of consistency is based on the following . . .

EXTSD = sd(e1 to e10).

AGRSD = sd(a1 to a10).

CONSD = sd(c1 to c10).

STASD = sd(s1 to s10).

OPNSD = sd(o1 to o10).

V = mean(EXTSD to OPNSD).

That is, we compute the standard deviation of responses to the items within each Big Five dimension. Each standard deviation is a measure of inconsistency in responding to the items within a dimension.

The overall measure of inconsistency, V, is the mean of the 5 within-dimension standard deviations.

V = 0 means no inconsistency – the respondent is perfectly consistent which would be weird for humans.

Values of V larger than 0 represent varying amounts of inconsistency.

The regression

The dependent variable is the variable called criterion. It is scores on the first midterm exam in PSY 1010.

The predictors are

con: conscientiousness scale scores

V: The measure of inconsistency computed above

wpt: Cognitive ability as measured by the Wonderlic Personnel test.

vXcon: The product of V and conscientiousness, used for the test of moderated regression, i.e., the interaction of V and con.

Your task

1. Compute V using the information above.

2. Assess the significance of wpt, con, and V predicting criterion in a multiple regression conducted using GLM.

3. Compute the product of V and con. Call it vXcon.

4. Assess the significance of wpt, con, V, and vXcon in a multiple regression using GLM.

If the product term is significant, that means that V moderates the con->criterion relationship. Moderation means that the con --> criterion relationship is different for different values of V.

5. Create graphs to display the moderation (or lack of moderation)

Form a variable called vgroup which = 0 if V is less than or equal to the median of all V values and = 1 if V is larger than the median of all V values.

Compute the correlation of criterion with con for vgroup=0 and for vgroup = 1. (Hint – use SPSS’s Data  Select Cases pull down to select either vgroup=0 or vgroup=1 for the separate correlations.)

Submit . . .

NUMBER EACH PART AS I’VE DONE BELOW AND BEGIN EACH PART ON A NEW PAGE.

1. Histogram and mean and SD of V to show that you got it correct.

2. GLM “Tests of Between-Subjects Effects” and “Parameter Estimates” tables to show that you did the prediction correctly for the Con+wpt+V equation.

3. Histogram of vXcon along with mean and SD to show that you computed it correctly.

4. GLM “Tests of Between-Subjects Effects” and “Parameter Estimates” tables to show that you did the moderated regression correctly for the Con+wpt+V+VxCon equation.

5. The 2 graphs – one for the V group=0 and one for V Group=1.

6. A 1-2 paragraph written statement describing what you found and what this might mean in selection of employees, for example.