help > Interpreting ANCOVA and multiple regression
Showing 1-1 of 1 posts
Jan 9, 2017 11:01 AM | Alice Yo
Interpreting ANCOVA and multiple regression
Dear Alfonso
I have two questions regarding multiple regression and ANCOVA models and I would appreciate your help:
1) ANCOVA:
I have 2 groups of subjects, 4 clinical scales with scores for the two groups, and a 5th additional scale that is only relevant for one group and has scores only for that group.
- When looking at between group differences using two-sample t-tests without accounting for any scales, there are wide spread ROI-ROI functional connections that are stronger in one group.
- When testing the between group differences in a one way ANCOVA model with the 4 clinical scales above and including interaction terms (i.e., including 8 covariates for the 4 clinical scales) all the between group differences disappear.
- Alternatively, when testing the between group differences in a one way ANCOVA model now including only the 5th scale (not included in the above analysis) all the between group differences also disappear.
So, I was wondering how to interpret these results. Is it right to state that all the between-group differences can be explained either by the 5th scale or the group of the 4 scales?
2) Multiple regression:
I have one group of subjects and 5 clinical scales and I test a multiple regression model to predict ROI-ROI functional connections for each clinical scale when accounting for the other 4 scales. There is a set of functional connections predicted by a certain clinical scale (S1). Now, when I replace S1 in the model with a scale R1 that has a Pearson's correlation of 0.7 with S1, and keep all the other original 4 scales in the model, there is now no significant association of functional connections with R1. I was wondering if this makes sense due to the high correlation between S1,R1.
Thank you very much for your help and this great forum,
Alice
I have two questions regarding multiple regression and ANCOVA models and I would appreciate your help:
1) ANCOVA:
I have 2 groups of subjects, 4 clinical scales with scores for the two groups, and a 5th additional scale that is only relevant for one group and has scores only for that group.
- When looking at between group differences using two-sample t-tests without accounting for any scales, there are wide spread ROI-ROI functional connections that are stronger in one group.
- When testing the between group differences in a one way ANCOVA model with the 4 clinical scales above and including interaction terms (i.e., including 8 covariates for the 4 clinical scales) all the between group differences disappear.
- Alternatively, when testing the between group differences in a one way ANCOVA model now including only the 5th scale (not included in the above analysis) all the between group differences also disappear.
So, I was wondering how to interpret these results. Is it right to state that all the between-group differences can be explained either by the 5th scale or the group of the 4 scales?
2) Multiple regression:
I have one group of subjects and 5 clinical scales and I test a multiple regression model to predict ROI-ROI functional connections for each clinical scale when accounting for the other 4 scales. There is a set of functional connections predicted by a certain clinical scale (S1). Now, when I replace S1 in the model with a scale R1 that has a Pearson's correlation of 0.7 with S1, and keep all the other original 4 scales in the model, there is now no significant association of functional connections with R1. I was wondering if this makes sense due to the high correlation between S1,R1.
Thank you very much for your help and this great forum,
Alice