small sample correction and missing data

[Anouk van der Straten]

Dear SwE experts,

I have done an fMRI experiment looking at the neuroimaging effects of different types of treatment (medication vs. psychotherapy) in patients with obsessive-compulsive disorder. I have three groups (OCD medication, OCD psychotherapy, healthy controls) and two timepoints (T0 = before treatment, T1 = after treatment (or after 16 weeks in HC). 

My groups:

1. 21 OCD patients treated with medication (of which 18 have completed the second session)

2. 24 OCD patients treated with psychotherapy (of which 16 have completed the second session)

3. 20 healthy controls (which all have completed the second session)

I want to look at two different F-contrasts:

- Group (OCD vs. HC) x time interaction

- Group (medication vs. psychotherapy) x time interaction (corrected with a covariate for the respons rate)

I have two questions:

1. Do you advice to perform a small sample adjustment or not with the size of these groups? In addition, I have read in previous conversations that if the n is smaller then 25, you advise against specifing different groups (but make the distinction with the covariates). Would this also be the case for my design?

2. I have now run my model including all the patients (also the ones that have only completed the first session) to correct for a possible drop-out bias in both patient groups. Is this the correct way to approach this analysis and is it valid with the number of participants that dropped out compared to the whole group?

Thank you for your help!

Kind regards,

Anouk van der Straten