Access the data file "regressmediation.sav" on Moodle.  File includes:

• sex
• age
• optimism - 6 items; e.g., “In uncertain times I usually expect the best” 1-5 Likert scale
• perceived stress - 5 items; “To what extent have you felt this way in the past few weeks:” e.g., active irritable, nervous; 1 (very slightly or not at all) – 5 (extremely)
• life satisfaction - 5 items; e.g., “The conditions of my life are excellent.” 1 – 7 Likert
• perceived control of internal states - 18 items; a measure of the degree to which people feel they have control of their internal states (emotions, thoughts, physical reactions). 1-5 Likert.
  

1.  Although the optimism, stress and control scales have already been created for you, the life satisfaction scale has not.  Create the life satisfaction scale (average the items). None of the 5 items needs to be recoded.

2. Compute the bivariate correlations amongst the variables perceived control, life satisfaction, stress and age.  Write down the correlations.

3. Age and stress are significantly correlated. Look for other variables that correlate with age and with stress. Based on these other correlations, why might the bivariate correlation between age and stress be misleading? Draw Venn diagrams to explain your reasoning. (Note - you do not need to discuss mediation to answer this question. Instead, consider the more basic idea of overlapping variance.)

4. Conduct a single multiple regression analysis with the variables from #3. Stress should be the outcome variable. Which variable(s) are the best predictor(s) of stress? Why is the relationship between age and stress different in this analysis compared to the bivariate correlation you previously computed? Write a brief paragraph answering these questions. Provide statistical evidence where relevant.

To report a regression analysis:

• First the “overall equation” as significant or non-significant; R= .xx, F(x ,xxx) = x.xx, p = .xx.
• Then report the Betas and significance for each predictor; Beta = .xx, t = x.xx, p = .xx.

5. Look under “Data” and “Split File.” Click “organize output by groups” and move the variable “sex” into the dialog box. Then run correlations between age and optimism. Does sex appear to moderate the relationship between age and optimism? Provide the statistical information which supports your answer.

*Note: Don’t forget to turn “split file” off.

6. Type out the formal way to phrase this mediational hypothesis: We hypothesize that perceived control leads to greater life satisfaction because of the reduction in stress caused by perceived control. (i.e., what mediates which relationship?)

7. Do a path analysis (i.e., run a series of regression analyses) to test the mediational hypothesis from #5.

8.  Re-type the paragraph below, filling in the missing information.  Also, create an APA Style table reporting on the bivariate correlations you computed in #2.  (You only need to report the correlations for life satisfaction, stress and perceived control.)   See this PDF document for a model. You should label this, "Table 1."

We noted that there were significant bivariate correlations amongst all of our variables (see Table 1), and predicted that (fill in your response for #6). This mediational hypothesis (was/was not) supported. The predictor variable (X) was significantly related to both the proposed mediator (Z; R=.xx, F(df, df) = X.xx, p = .xx) and the outcome variable (Y); R=.xx, F(df, df) = X.xx, p = .xx. Additionally, Z was significantly related to Y; R=.xx, F(df, df) = X.xx, p = .xx . To test for mediation, we conducted a (name analysis) and entered X and Z as predictor variables and Y as the outcome variable. The overall equation was significant; R=.xx, F(df, df) = X.xx, p = .xx . Z’s relationship with Y remained significant even while controlling for X; Beta = .xx; t = x.xx, p = .xx. Most importantly, the relationship between X and Y was weaker in this analysis (Beta = .xx; t = x.xx, p = .xx) compared to the direct relationship (Beta = ). These results suggest (partial/full) mediation.