Colleagues and I ran an experiment which gives us a set of correlations. We preregistered at the Open Science Framework at https://osf.io/593sz/registrations/. Preregistration involves specifying any transformations to be performed on the data. In the absence of any theoretical predictions or pilot data, I designed a procedure that would pick the transformation for each data set. I ended up with most of the data being squared, making me wonder whether I get higher correlation coefficients only because the analysis has become more sensitive to outliers. So is there a procedure that tells me how much a correlation coefficient depends on outliers? I don't want to identify individual data points that can be considered outliers, I want to know how much the observed r depends on outliers. I have failed to find anything, and the colleague we usually consult when we have such questions couldn't think of anything, either.