Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
| From | Nick Cox <njcoxstata@gmail.com> |
| To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
| Subject | Re: st: RE: Re: estout |
| Date | Wed, 12 Mar 2014 09:12:21 +0000 |
Thanks for providing a reproducible example. As I said, I don't use -estout- (SSC) but I do recall some properties of exp(). Setting aside what the numbers represent, the difference between any exp(x_1) and any exp(x_2) can be positive, zero, or negative, so there is no surprise that it can be negative. A fortiori, what you are calculating cannot be a standard error. The error here is not in -estout- and not in Stata [NB] but, I guess, in using an incorrect formula. Nick njcoxstata@gmail.com On 12 March 2014 04:46, Yuesheng Wu <wuyuesheng20@gmail.com> wrote: > Hi Nick, > That is a very good point; here is the example I can come up with: > sysuse auto > poisson price weight mpg > > estout, cells(b se) transform(weight > exp(_b[weight]+_b[mpg])-exp(_b[weight]) > exp(_b[weight]+_b[mpg])-exp(_b[weight]) mpg > exp(_b[weight]+_b[mpg])-exp(_b[weight]) exp(_b[weight]+_b[mpg])) > > > Here, I want to calculate the term > "exp(_b[weight]+_b[mpg])-exp(_b[weight]) "; the STATA gives the > following results: > > b/se > price > weight -.0123634 > -1.000255 > mpg -.0123634 > .0004759 > _cons 8.179377 > .0194376 > > Note that the standard error is -1.000255. > > Any thoughts? > > Thanks a lot! > > Yuesheng > > On Tue, Mar 11, 2014 at 2:18 PM, Nick Cox <njcoxstata@gmail.com> wrote: >> No doubt, but you are still not giving any details of the code you used. >> >> How can users of -estout- (SSC) (not me, by the way) explain whether >> you messed up your code when you don't even say what it is! >> >> A reproducible example -- one based on a dataset we can all access -- >> would be even better. >> >> Nick >> njcoxstata@gmail.com >> >> Yuesheng Wu >> >>> Thanks for your help. >>> I did not export the estimation results to Excel: I just did >>> everything in Stata. >> >> On Tue, Mar 11, 2014 at 1:32 PM, Rubil Ivica <irubil@eizg.hr> wrote: >> >>>> you are probably exporting your results to Excel, right? If so, then you >>>> need either of the following two options: "nopa" or "brackets". >>>> The problem is that estout by default puts standard errors in >>>> parentheses (std.err.) and when this is exported to Excel, it becomes >>>> -std.err. This >>>> is a convention in accounting that instead od negative values those are >>>> put in parentheses. "nopa" will drop parentheses, and "brackets" will >>>> put >>>> berackets [...] instead of parentheses (...). Hopefully this helps. >> >> Yuesheng Wu >> >>>> After I run the Poisson regression by "poisson", I use to estout to >>>> calculate standard error of a term with the estimates of two parameters; >>>> it turns out the standard error is negative. I am just wondering whether >>>> the estout only works for one parameter; or I mess up with the code. >>>> >> * >> * For searches and help try: >> * http://www.stata.com/help.cgi?search >> * http://www.stata.com/support/faqs/resources/statalist-faq/ >> * http://www.ats.ucla.edu/stat/stata/ > * > * For searches and help try: > * http://www.stata.com/help.cgi?search > * http://www.stata.com/support/faqs/resources/statalist-faq/ > * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/