unequal sample sizes

[skushner]

I'm hoping for help on dealing with unequal sample sizes. Any assistance is greatly appreciated.

I'm currently working with data from schizophrenic, bipolar and depressed patients. We will be comparing these samples to determine whether these diagnostic groups can be differentiated based on mean scale scores from a commonly used personality inventory. We will be comparing mean scores using MANOVAs and classification accuracy using ROC curves and descriptive efficiency statistics.

I am unclear how to deal with the unequal size of the patient samples: currently, our data set has 70 schizophrenic, 70 bipolar and 423 depressed patients (this data has been accumulated through affiliation with several clinics and research studies). Is it necessary to select a random sample from the group of depressed patients? Are there other techniques I should conduct? Or, can I proceed with the sample sizes as is?

Thanks!

[CowboyBear]

I am unclear how to deal with the unequal size of the patient samples: currently, our data set has 70 schizophrenic, 70 bipolar and 423 depressed patients (this data has been accumulated through affiliation with several clinics and research studies). Is it necessary to select a random sample from the group of depressed patients? Are there other techniques I should conduct? Or, can I proceed with the sample sizes as is?

Hi there. Having unequal cell sizes may reduce the robustness of your MANOVA to assumption violations, especially heterogeneity of variance-covariance matrices. Your statistical power is obviously also limited by the size of the smaller groups.

Cutting the depressed sample down to 70 might help the robustness aspect, but you're also likely to lose a fair bit of information by doing so, and reduce the degree to which you can generalise the results to other depressed people. I might suggest only doing this if the variance-covariance hetereogeneity is a problem. Check via Levene's test and Box's m (there are others that I don't recall offhand, sorry!)

Another technique you could utilise if you don't feel you can justify the assumptions of MANOVA is multinomial logistic regression, an unfairly ignored technique in psychology

By the way - I must say I'm wondering (I'm a psychology student/researcher myself) - what is the theoretical bent behind your research? As far as I know the association between disorders s/a schizophrenia/bipolar/MDD and common personality tests has been researched pretty extensively already?

[skushner]

Thanks for the suggestion. We are investigating the most recent revision of a measure, which has added new scales, and has not been extensively investigated with respect to differential diagnosis.

[CowboyBear]

Ah, I see, cool! Good luck with your study