ANOVA results and confidence intervals

#1
Hallo can someone help me with this problem, please?

I have done a univariate 2 factor ANOVA with spss.
The problem is that the 95% confidence intervall for my mean values (shown in description table) are overlaping each another - and therefore I suppose are not sign. But ANOVA says that the factors have sign effects (testing on interactions was not sign).

I don't know how spss calculates the confidence intervals, but I cannot say the groups have sign means according to ANOVA when my bar chart shows overlapping confidence intervals.

Thanks for help.
 

JohnM

TS Contributor
#2
It is possible to have a significant F statistic in the overall ANOVA, but if the factors have >2 "levels," when the means are compared as in a post-hoc analysis, the differences between them may not be significant.

I would need to see your ANOVA table and an image of the graph in order to verify that.
 

JohnM

TS Contributor
#3
sirhoppel said:
Hallo can someone help me with this problem, please?

I have done a univariate 2 factor ANOVA with spss.
The problem is that the 95% confidence intervall for my mean values (shown in description table) are overlaping each another - and therefore I suppose are not sign. But ANOVA says that the factors have sign effects (testing on interactions was not sign).

I don't know how spss calculates the confidence intervals, but I cannot say the groups have sign means according to ANOVA when my bar chart shows overlapping confidence intervals.

Thanks for help.
I would conclude that the specific differences between the means (with overlapping confidence intervals) are not significant - if part of your study was to examine an overall effect, then that would be significant.