I have a small data set that consists solely of counts. I have several variables. For example State, for each state (Mexican States), for example, Tamauplipas I have the counts of domestic violence and the counts of non-violence. For Age I have the same but for each age group... I would eventually have a table, for example, like this:


I naturally have many more columns. What these people want to know (plot) is what levels of what variable are associated with Non-Violent and the same for Violent. I was told in work to do a biplot but have been trying and dont get nice results. I'm not sure this is the best idea. I tried with a regular PCA biplot... What do you guys think? any comments would be appreciated. Greetings!



1 Answer 1


A biplot only works if the first two principal components account for most of the variance in your data. Otherwise, you won't get useful results. You might not be able to plot your data in two dimensions.

If I'm looking at your data set correctly, you'd only have two points on your biplot - one point for violent and one point for non violent. Are there only two rows in your data set? If so, you're not going to get anything out of a biplot or PCA in general.

All I can think of for count data like that would be some barplots, particularly the stacked bar plots seen here:


  • $\begingroup$ Agreed here, but I was thinking that - if indeed there are only two rows - let one axis be "violent" let the other axis be "non-violent" and plot each column in this 2-d space as a scatter plot. With only two components, there's no dimensionality in this dataset to reduce. Maybe centering/scaling each row with improve the plot. $\endgroup$
    – zzk
    Oct 2, 2012 at 22:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.