Skip to main content
11 events
when toggle format what by license comment
Dec 9, 2020 at 8:11 vote accept Richard Hardy
Dec 3, 2020 at 19:39 comment added Alecos Papadopoulos @CloseToc If we really cannot do anything about it, still, it is important to know it, because we should know that, for example, the estimator variance will be wrongly computed. In practice, in such instances researchers may attempt a latent- or proxy- variable model, linking the unavailable $m$ to some available data.
Dec 3, 2020 at 19:19 comment added CloseToC It seems like the modelling-relevant instances of unconditional heteroscedasticity are really instances of conditional heteroscedasticity: There is some observable variable, with which which the error variance has a systematic relationship ($m$, or a proxy). I wonder whether anything is to be gained from entertaining the notion that error variances could vary across an index when there is no such variable.
Dec 3, 2020 at 14:03 history edited Alecos Papadopoulos CC BY-SA 4.0
deleted 5 characters in body
Dec 3, 2020 at 13:47 history edited Alecos Papadopoulos CC BY-SA 4.0
Added example responding to comments
Dec 3, 2020 at 10:17 comment added Henry and give an example of "unconditional heteroskedasticity"
Dec 3, 2020 at 8:40 comment added Richard Hardy Could you outline briefly what you mean by structural reasoning?
Dec 2, 2020 at 20:24 history edited Alecos Papadopoulos CC BY-SA 4.0
deleted 1 character in body
Dec 2, 2020 at 19:30 comment added Alecos Papadopoulos @whuber By structural reasoning. At least this is what we do in economics.
Dec 2, 2020 at 19:26 comment added whuber If we don't have variables on which the amount of scatter depends, how are we even to recognize heteroscedasticity?
Dec 2, 2020 at 19:20 history answered Alecos Papadopoulos CC BY-SA 4.0