[SOLVED] statistic network theory Regression and Multivariate Statistics

Regression and Multivariate Statistics

Exercise #4

This exercise makes use of the miscarriage data set that was described in previous exercises.Use the data set EX4.SAV (EX4.xls, ex4.rdata), which is available in the same folder.When the women were interviewed at six weeks post miscarriage, they were asked about the level of perceived social support.The measure (PERCSS) ranges from 0 (no perception of support) to 4 (high levels of perceived support).Because all these women are experiencing the stress of the miscarriage, we expect that on average, perceived support should be negatively related to symptom level at followup.Note that the perceived social support measure was at the end of the interview, and it is often missing.

Clearly all women in this study had sexual partners, but one third of them are not married.The unmarried women describe themselves as single, steadily attached, or divorced/separated.One expects that the women who are married are, on the average, likely to get some support from their spouse, and therefore that their symptoms will be lower at followup.The data set has a variable MARRIED which is 1 for married, 0 otherwise.We consider this variable to be one proxy for network support.

One perspective argues that perceived social support will be especially important for women who have less committed partners, i.e. those who are unmarried.We will test this hypothesis this week.

Regress CESD2 (CESD at six weeks post miscarriage, without transformation) on CESD1 (step 1), MARRIED & PERCSS (step 2), and the interaction of MARRIED and PERCSS.You will have to create a new variable to represent the interaction.We recommend that PERCSS be centered before creating the new variable.

The models will thus be

1) CESD2 ~ CESD1
2) CESD2 ~ CESD1 + MAR + PER
3) CESD2 ~ CESD1 + MAR + PER + MAR*PER

Is there evidence that the two support variables (MARRIED and PERCSS) are related to 6 week CESD when 2 week CESD is adjusted?

Evaluate the evidence that marital status moderates the effect of PERCSS.Consider both standard significance tests and the confidence bound on the interaction parameter.

Regardless of the conclusion regarding part 1b, write a statement that interprets all of the coefficients from the full moderation model.

Suppose that an alternate theory suggests that the effect of MARRIED is mediated by perceived social support (PERCSS).

List each of the models that have to be run in order to evaluate this theory.

Run each of the models and determine if there is a meaningful mediation effect of PERCSS on MARRIED.

Calculate the indirect effect, its standard error, and Sobel Z statistic.Determine if the indirect effect is significantly different from zero (the null hypothesis).Calculate a 95% CI for the indirect effect.

Describe the problem with the Sobel Z-test and calculate a bootstrap estimate for the indirect effect.Is the bootstrap result consistent with the Sobel Z-test? Compare the 95% CI for the bootstrap results with the 95% CI from the Sobel method.