Hi Alfonso,
I am a little confused about setting up age as a covariate in analyses.
For example, if I have a 2 group comparison (treatment, control) on some outcome and I want to control for age, I can enter [1 -1 0], and this controls for age overall (effectively like a covariate/control variable in regression)?
If I wanted to control for age and any age-by-group interactions, I would create separate variables for each group, with ages entered for the subjects in that group, and zeros for the rest e.g. for groups of n=5: treatment_age 35 45 55 61 78 0 0 0 0 0 control_age 0 0 0 0 0 21 33 44 55 43
And then enter a contrast of [1 -1 0 0], wth the zeroes reflecting the treatment_age and control_age groups.
However, if I wanted to control for the effect of age overall AND the potential age-by-group interaction would I need to enter age, treatment_age, and control_age into the model i.e. [1 -1 0 0 0]? Or does entering the age-by-group interaction also control for age overall and this would be redundant, and I should just use [1 -1 0 0].
Finally, would this controlling of age-by-group be considered a standard approach (e.g. for research looking at ageing and dementia), or just included one overall age control variable be sufficient i.e. the [1 -1 0] model?
Many thanks,
Andrew
I am a little confused about setting up age as a covariate in analyses.
For example, if I have a 2 group comparison (treatment, control) on some outcome and I want to control for age, I can enter [1 -1 0], and this controls for age overall (effectively like a covariate/control variable in regression)?
If I wanted to control for age and any age-by-group interactions, I would create separate variables for each group, with ages entered for the subjects in that group, and zeros for the rest e.g. for groups of n=5: treatment_age 35 45 55 61 78 0 0 0 0 0 control_age 0 0 0 0 0 21 33 44 55 43
And then enter a contrast of [1 -1 0 0], wth the zeroes reflecting the treatment_age and control_age groups.
However, if I wanted to control for the effect of age overall AND the potential age-by-group interaction would I need to enter age, treatment_age, and control_age into the model i.e. [1 -1 0 0 0]? Or does entering the age-by-group interaction also control for age overall and this would be redundant, and I should just use [1 -1 0 0].
Finally, would this controlling of age-by-group be considered a standard approach (e.g. for research looking at ageing and dementia), or just included one overall age control variable be sufficient i.e. the [1 -1 0] model?
Many thanks,
Andrew