help
help > RE: interaction effects conn
Jul 27, 2016 05:07 PM | Alfonso Nieto-Castanon - Boston University
RE: interaction effects conn
Dear Charlotte,
There are multiple equivalent ways to define those interaction analyses, but perhaps the simplest would be to have defined four covariates: case_female, case_male, control_female, and control_male (each containing 1's for the corresponding subgroup of subjects and 0's for everyone else). Then for the interaction analyses simply select all four effects and enter the contrast [1 -1 -1 1]. This analysis will show those areas where connectivity differences between males&females are different in the control group compared to the case group (ie. a case by gender interaction).
Regarding the additional IQ/age covariates, if you are planning to use these as covariates of no interest, for example in order to control for potential interaction effects that may be due to age and/or IQ differences between the groups, then yes, you would do that simply by selecting all 'age', 'IQ', 'case_female', 'case_male', 'control_female', and 'control_male' covariates, and entering a between-subjects contrast [0 0 1 -1 -1 1]. This analysis will show those areas where connectivity differences between males&females are different in the control group compared to the case group, after controlling for potential connectivity differences that can be explained by age or IQ differences between the same groups (ie. a case by gender interaction controlling for main age&IQ effects). If you want to see, in these same analyses, what was the main effect of age, for example, you would do that simply by entering instead the contrast [1 0 0 0 0 0] (or [0 1 0 0 0 0] if you want to look at the main effect of IQ), and that will show those areas where the connectivity is associated with age after controlling for potential case and gender effects.
Hope this helps
Alfonso
Originally posted by charlotte p:
There are multiple equivalent ways to define those interaction analyses, but perhaps the simplest would be to have defined four covariates: case_female, case_male, control_female, and control_male (each containing 1's for the corresponding subgroup of subjects and 0's for everyone else). Then for the interaction analyses simply select all four effects and enter the contrast [1 -1 -1 1]. This analysis will show those areas where connectivity differences between males&females are different in the control group compared to the case group (ie. a case by gender interaction).
Regarding the additional IQ/age covariates, if you are planning to use these as covariates of no interest, for example in order to control for potential interaction effects that may be due to age and/or IQ differences between the groups, then yes, you would do that simply by selecting all 'age', 'IQ', 'case_female', 'case_male', 'control_female', and 'control_male' covariates, and entering a between-subjects contrast [0 0 1 -1 -1 1]. This analysis will show those areas where connectivity differences between males&females are different in the control group compared to the case group, after controlling for potential connectivity differences that can be explained by age or IQ differences between the same groups (ie. a case by gender interaction controlling for main age&IQ effects). If you want to see, in these same analyses, what was the main effect of age, for example, you would do that simply by entering instead the contrast [1 0 0 0 0 0] (or [0 1 0 0 0 0] if you want to look at the main effect of IQ), and that will show those areas where the connectivity is associated with age after controlling for potential case and gender effects.
Hope this helps
Alfonso
Originally posted by charlotte p:
Dear Alfonso,
To investigate interaction effects in a 2 (case/control) by 2 (female/male) level design, am I correct in thinking that if I select my subject effects 'case', 'control', 'female', 'male' (each with a vector indicating 1 for yes, and 0 for no), entering [1 -1 -1 1] as the between-subject contrast will yield the interaction effect for (case > control) x (male > female)?
Further, to include further covariates like age or IQ, selecting one or more of those from the between-subjects factors, what does the contrast need to look like?
e.g. selected: age, IQ, case, control, female, male
what would I need to assess the effect of age on the interaction: [? ? 1 -1 -1 1]
what would I need to assess the effect of age and another covariate on the interaction?: [? ? 1 -1 -1 1]
Thank you very much for your help,
Best wishes,
Charlotte
To investigate interaction effects in a 2 (case/control) by 2 (female/male) level design, am I correct in thinking that if I select my subject effects 'case', 'control', 'female', 'male' (each with a vector indicating 1 for yes, and 0 for no), entering [1 -1 -1 1] as the between-subject contrast will yield the interaction effect for (case > control) x (male > female)?
Further, to include further covariates like age or IQ, selecting one or more of those from the between-subjects factors, what does the contrast need to look like?
e.g. selected: age, IQ, case, control, female, male
what would I need to assess the effect of age on the interaction: [? ? 1 -1 -1 1]
what would I need to assess the effect of age and another covariate on the interaction?: [? ? 1 -1 -1 1]
Thank you very much for your help,
Best wishes,
Charlotte
Threaded View
| Title | Author | Date |
|---|---|---|
| charlotte p | Jul 27, 2016 | |
| Alfonso Nieto-Castanon | Jul 27, 2016 | |
| charlotte p | Jul 27, 2016 | |