1

After doing a t-test that showed a significant difference, I did an ANCOVA with 3 covariates. The ANCOVA showed no significant effect of the variable that did show a significant difference in the t-test. However, one covariate showed a significant effect. This confused me a bit, so I have a few questions:

  1. What does it mean dat the significant difference found in the t-test is no longer significant when doing an ANCOVA?
  2. How to interpret the significant covariate in my ANCOVA?

enter image description here

user1008199
  • 11
  • 3
  • Is the covariate correlated with the grouping indicator? – gung - Reinstate Monica Jan 15 '16 at 15:04
  • The group is whether mothers conceived a child through assisted reproductive technology and the significant covariate is gender – user1008199 Jan 15 '16 at 15:06
  • So is gender associated with ART? – gung - Reinstate Monica Jan 15 '16 at 15:12
  • Yes, studies showed that ART boys had more emotional problems than ART girls. So I selected gender as a covariate. – user1008199 Jan 15 '16 at 15:17
  • I've added a screenshot of my table. So basically what I'm focusing at is ART on somatic complaints, which is showing no significant effect (it showed a significant difference in a t-test), while gender (covariate) is showing a significant effect. – user1008199 Jan 15 '16 at 15:25
  • This is just the difference between ignoring and controlling for another variable. You can get a better understanding of this issue from my answer [here](http://stats.stackexchange.com/a/78830/7290). – gung - Reinstate Monica Jan 15 '16 at 15:40
  • Thank you, I'm still not sure how to interpret these finding though. I cannot say there's a significant effect of ART on somatic complaints right? And could mention that gender has a statistically significant relationship with somatic complaints? Also, I'm not sure how to explain how there's a significant difference of ART on somatic complaints in the t-test while this disappeared in the ANCOVA? I had to pick up this work after quite some time, so it has been a while since I've done statistics. Sorry for the many questions. – user1008199 Jan 15 '16 at 15:51
  • It's in the linked thread. The reason ART appears to be associated is because it is associated w/ gender, & gender is associated w/ emotional problems. – gung - Reinstate Monica Jan 15 '16 at 15:53

0 Answers0