I would be grateful if people could give me their opinion on the following:
We have a cohort study which started 4 years ago in which 170 boys and a137 girls were randomly selected from a number of urban schools (stratified by SES). It was originally thought that both genders could be analysed together, and the total of 307 children gave us much more than the 80% power required for such studies.
Apx 30 children have dropped out since the beginning of the study (no gender diff, no SES diff). Moreover, since we have come to realise that boys and girls differ significantly in most outcome measures. We now carry our analyses separately for each gender. We are therefore concerned about a possible reduction of power on certain measures, especially in the girls.
We have the opportunity to recruit apx 40 new children. Is there anything wrong in recruiting only girls? We would solve our power problems. One concern is that we might be introducing some kind of bias. Would it make more sense to recruit 20 boys and 20 girls? In this case we would not increase our power enough in girls...
Thank you for any comments, I look forward to receiving your replies
Kind Regards
Sandra
|