Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Austin Nichols <austinnichols@gmail.com> |
To | statalist@hsphsun2.harvard.edu |
Subject | Re: st: RE: weighting prior to creating cook's distance stats for reg |
Date | Tue, 14 Dec 2010 21:43:24 -0500 |
See also http://www.stata.com/meeting/dcconf09/dc09_valliant.pdf On Tue, Dec 14, 2010 at 1:04 PM, Nick Cox <n.j.cox@durham.ac.uk> wrote: > Some experimentation suggests that Cook's [sic] distance is not available with any kind of weights other than fweights. > > I agree that the error message you got is puzzling. > > It seems that you would need to define your own version of Cook's distance with pweights (I've no idea how much sense that makes) and your own guidelines on how to interpret the results -- unless of course you can cite literature on this. > > Program writers have various stopping rules, including > > 1. Product must ship soon. > > 2. No idea if that is defined theoretically. > > 3. Not clear if that is more help than hindrance. > > But I can't speak for which of these, or of others, was at work here. > > Thanks in advance for not using the formulaic "Thanks, in advance". > > Nick > n.j.cox@durham.ac.uk > > Ben Hoen > > I am estimating a weighted regression (using the weights generated from a > coarsened exact matching (cem) procedure (see > http://gking.harvard.edu/cem/)). > > .reg lasp ee sqft acre age i.nbrhd [pweight=cem_weight] > > where: > lasp = log of sale price > ee = if home has energy efficiency features > sqft = square feet of the home > acre = acres of the parcel > age = age of the home at sale > nbrhd = the neighborhood in which the home resides > > An important post-estimation stat I would like to create is cook's distance > (cdis). > > I am receiving the following errors when I try to estimate them: > > . predict cdis if e(sample), cooksd > option cooksd not allowed after robust estimation > r(198); > > I am not using the vce(r) option and therefore cannot understand the reason > for the error. More importantly, is there a way to produce cook's distance > stats after the weighted regression I estimated? > * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/