Deutsches R-Forum

I don't know how to interprete this plot.

Help would be appreciated! Thanks!

It shows fitted values vs residuals, likely from a gaussian linear model.

We see distinct lines, indicating that y is discrete (?), e.g. counts?

The lower border indicates that y has a lower bound (e.g. counts or positive only data.

The fan shape indicates a positive mean-variance relationship.

All three indicate that the gaussian linear model might be not a good representation.

A log-transformation of y could fix this.

I'd prefer a GLM with Gamma (for positive continuous y) or poisson / negative-binomial (depending on mean-variance-relationship) distribution. In both cases likely with a log-link function.

HTH,

Edi