I am currently working on my MPH thesis and have come across an issue. I am using percentages as predictor variables (e.g., percentage of children taken to health care provider, completeness of birth registration). Can I use Pearson´s correlation coefficient and perform simple linear regression with these data? I apologise if this is a stupid question, but I am a real novice here...
Thank you, and best wishes,
i dont think it's a stupid question but a very valid one.
it really depends on how much variability you have across the ranges of percentage scores (among other things). you could end up with negative percentages in the fitted values depending on how they are but if you have a good spread of them i dont really see a problem wiht using them in regression
Thanks, Oscar, this is very helpful.
Since you use percentages as predictors, not the outcome, you do not run the risk of having negative predicted percentages. As far as I know, having limited independent variables (percentages are within 0 and 100) is not a problem for correlations or regressions.
that's a good point Ingo. i dunno why i assumed the outcome variables were also percentages but afre re-reading the OP's post i realised that you're right. we generally don't care too much about the range of the predictors.
Thank you to both of you, I really appreciate that.