Reporting Dummy Variables and Determining Significance

#1
Hi all,

I was just wondering, how does a person determine the significance of a categorical variable in a hierarchical regression model if one dummy variable is significant and the other is non-significant?

So basically, the categorical variable (X) I am examining has three levels, so therefore I created two dummy variables with 'A' as the baseline category: A vs. B and A vs. C. I entered the two dummy variables alongside three scale variables in a hierarchical regression model at Step 1, while there were two further scale variables entered at Step 2. At Step 1, both dummy variables were significant predictors of my outcome Y. However, at Step 2, the addition of further variables rendered one of the dummy variables non-significant.

How would I report the significance in the Results section of my report? Would I say something along the lines of:

At Step 1 X was a significant predictor of Y. However, X was rendered non-significant at Step 2 when further variables were taken into account.

If one dummy variable is rendered non-signficant, does that mean that the variable as a whole is not a significant predictor of a given outcome? In other words, is it the case that all dummy variables have to be significant in order for the variable as a whole to be considered a significant predictor? I've looked through several statistics books, but there doesn't seem to be a definitive answer on the matter.
 

hlsmith

Omega Contributor
#2
Two things jump into my mind. You can perform a -2loglikelihood test perhaps or if the model allows, you may be able to look at type III coefficient effects.