How heterogeneous is my sample?

[Rob Wood]

Hi there,

I have a data set of patients receiving a particular treatment. I have been asked to investigate whether or not there exists some measure of heterogeneity, such as an index or a score, that will indicate how heterogeneous my sample of patients are. All I could think of to use initially were measures of variation for each variable within the data set.

Any alternative ideas or help would be appreciated.

Thanks,

Rob.

[Richard Williams]

My inclination is the same as yours. In addition you may want to compare your sample to other samples or populations. People will often say that other samples are flawed or results have limited generalizability. So, for example, if other studies have primarily looked at women, whites, college students, rich people, poor people, or whatever, you might be able to cite the greater diversity of your sample as a strength. Conversely, if your sample is more homogeneous than others have been, maybe you can argue that this provides greater control over confounding influences. I suppose somebody may have created a hetero scale, but personally I would want to see the amount of diversity on several major demographic variables. Making comparisons to reference populations/ samples will help to clarify how diverse your sample is.

[Rob Wood]

Thank you for your thoughts Richard, there were very insightful, and I believe comparisons to other data sources may be the key!

Thanks again.