# Comparing groups within specific conditions

So I have a hypothesis that goes as follows: the relationship between gender of the leader and perceived effectiveness is moderated by company performance. I did some analyses on this and this moderation is significant.

The point is, however, I need to figure out whether female leaders in high company performance are perceived to be less effective than male leaders. In other words, is there a significant difference between the effectiveness of males and females in the SPECIFIC group of high company performance. I can't for the life of me figure out how to do this in SPSS so any pointers would be greatly appreciated.

• At least in the social sciences, you'd follow up a significant moderation with simple slopes analysis - that is, is the effect of gender on perceived effectiveness significantly different from 0 when company performance of 1 SD above the mean? This might be helpful for performing that in SPSS: psych-it.com.au/Psychlopedia/article.asp?id=259 Commented Oct 5, 2014 at 6:16

It's a long time since I used SPSS, so I am unsure of how you'd specifically achieve it in SPSS, but I'd probably be looking to set up a contrast that represents this specific difference and test that.

For example, such a contrast might be set up like so:

    Group                  Contrast
(Sex x Performance)          code

Female.High                   1
Male.High                    -1
Female.Low                    0
Male.Low                      0


You could then do a one-tailed test: If the coefficient of the contrast is significantly smaller than 0, the conclusion would be that female leaders in high company performance are perceived to be less effective than male leaders.

• Dear Glen, how about other combinations, say for example you have a situation like this one, can we compare A1-B1 (A1 minus B1) with A2-B2 (A2 minus B2)? Commented Apr 16, 2021 at 19:58
• Sounds like a whole new question, not a clarification of this answer, but wouldn't it just be a test of the interaction term? Commented Apr 17, 2021 at 14:04
• So, can we specifically test two different arrangements of interaction, namely: (1) A1-B1 versus A2-B1, and (2) A1-B1 versus A2-B2? BTW, my question is HERE. Commented Apr 17, 2021 at 15:53