Hypothesis test on quantile slope against OLS slope

Question

Using the excellent quantreg package and testing hypothesis of equal slopes between quantile 0.1 and 0.9 using standard anova function.

model <- food ~ income anova(rq(model,tau=0.1), rq(model,tau=0.9))

A zero p-val leads us to reject the null hypothesis that the slopes of q1 and q9 are equal

When running the same anova to compare quantile regression against OLS, there is an error

anova(rq(model,tau=0.5), lm(model))

Any suggestions?

Check `?anova.rq`, it seems that you can use only `rq` objects for comparison. — Julius Vainora, Nov 29 '15 at 20:42
@Julius. Indeed; we're trying to find solution for anova on both rq and lm models. — rrg, Nov 30 '15 at 04:51

score 0 · Answer 1 · answered Nov 29 '15 at 21:16

Testing with anova.rq only understands rq objects, in principle one could extend this to lm objects, I suppose, but things get a bit complicated. These notes meanmed by Tom Ferguson explain some details. In the regression setting things are a bit worse since one has local densities rather than one global one.

Hypothesis test on quantile slope against OLS slope

1 Answers1