quantile regression question

[startz]

Out of curiousity, why is it that when you do a quantile (median) regression you get a different estimate each time? For example,

Code: Select all

create u 1000 series x = rnd ls x c

Change the estimation to quantile and then repeatedly hit estimate. The estimates change each time by roughly a standard error.

[EViews Glenn]

I see no changes in the estimates when estimating the same equation on the same data.

[startz]

I can't either...although I was able to reproduce this any number of times earlier in the day.

[EViews Gareth]

You weren't repeatedly running that program, were you?

[startz]

No. Just hitting Estimate again. You think I kept generating new random numbers? How du... Actually, I don't think I want to finish that question. :)

[EViews Glenn]

My original response to your first post asked about the possibility that you were re-generating data. But I changed my response after giving you the benefit of the doubt.