My PC1 values are the negative of what I got last week (or what R / another library gives). Which one is correct?
Why this happens
A principal component is a direction, and a direction can point either way. If v is a component, then -v describes exactly the same axis and explains exactly the same variance.
Different solvers, library versions, or even small changes in the data can return either one. Both are correct.
Answer 1: Understand that the sign doesn't change the meaning
Distances, explained variance, and model performance are all unaffected by a sign flip. Only the interpretation changes. "High PC1 = wealthy" becomes "low PC1 = wealthy". If your downstream model is retrained on the new output, the flip makes no difference.
