I'm carrying out an exploratory PCA, followed by an CFA, to test for dimensionality of a questionnaire. I am wondering about the effect of having family members in the same sample.

I have a final sample of 1060 in total. This is made up of 530 autistic men (39 of whom are parents) and 530 autistic women (39 of whom are parents), with each sample matched approximately by age. It is fairly likely that the children of these parents also completed the questionnaire and would be included in the sample, although there is no way of knowing which respondents these are.

Should I remove the parents because the common home environment may lead to specific shared interests/lack of interests, thereby contributing to common variance in the factor analysis? I thought that this would be minimal as they would only be in clusters of twos or threes or fours, and it would be a shame to eliminate all 78 participants.

I'd really appreciate any help, thank you so much!


  • $\begingroup$ Interesting. But all you've said about the content is the indirect mention of "interests/lack of interests." Please say more about what's been asked of your participants. Also, to which population do you want to make inferences--that of non-parents alone? $\endgroup$ – rolando2 Mar 30 at 16:00
  • $\begingroup$ The questionnaire is designed to measure 'systemising' - the drive to construct systems (e.g. mechanic/ linguistic / natural). I would say it's likely, therefore, that an autistic mother or father that is interested in systems such as cars/ planes/ stock markets/ weather patterns/ etc would influence their autistic child to be interested in the same things. $\endgroup$ – T August Mar 30 at 16:21
  • $\begingroup$ And I am looking to draw conclusions about Autistic men and women generally. Specifically, I want to see whether their answers cluster along two or three factors of "systemising", or whether systemising ability can be considered unidimensional. $\endgroup$ – T August Mar 30 at 16:24
  • $\begingroup$ It seems there is no way to make the desired inferences about all autistic men and women without either excluding the parents among them or including the children. Maybe you'll want to try it both ways to see how or to what extent results differ. But I hope you will conduct the CFA on a different sub-sample than used for the EFA. Otherwise it's circular; or you might say like shooting fish in a barrel. $\endgroup$ – rolando2 Mar 30 at 19:26
  • $\begingroup$ Hello, yes I am splitting the samples and carrying out EFA and CFA separately! Actually on that point I actually have a second question... I was going to randomly mix the total samples and then split it... but am I better off also matching these groups by sex/age? $\endgroup$ – T August Mar 31 at 7:59

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.