18

separated into dummy variables. This number of independent variables

complicated the analysis to a significant degree. Therefore, in order to limit the

number of variables, we used a method of analysis that involved doing separate

regressions for each of the dimensions of interest (demographic characteristics,

negativity, deviance, and intimacy) and using the significant variables in each of

those regressions in a final regression. Because the variables had been pre-

selected for their level of significance, they were held to a far more conservative

test of significance in the final analysis. The significance test will be explained in

detail in the section that follows.

Because the two dependent variables, negative affect and positive affect,

have opposite valences, making sense out of the signs of coefficients can be

complicated. For ease of interpretation, the dependent variable, “negative affect”

was multiplied by –1. As a result, a negative coefficient for either negative or

positive affect can be interpreted as the respondent experiencing more negative

affect.

**Final Regressions **

We included all variables that were significant in the first four analyses as

the independent variables. The significance test used the critical z score

necessary for all of the combined variables in the equation to be significant as the

criteria for determining whether each individual variable was significant. This

was a far more conservative method than the more common t-test used in the

original equations. For the regression on negative affect, variables needed to have

a z-score of 3.0 or greater in order to be considered significant at the .05 level, a