Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | "P. Wilner Jeanty" <pwjeanty@gmail.com> |
To | statalist@hsphsun2.harvard.edu, "sanja_lutzeyer@ncsu.edu" <slutzey@ncsu.edu> |
Subject | Re: st: wtpcikr for models not estimated in STATA |
Date | Wed, 14 Nov 2012 13:56:49 -0600 |
On Wed, Nov 14, 2012 at 10:17 AM, sanja_lutzeyer@ncsu.edu <slutzey@ncsu.edu> wrote: > > Good day. > > I am using a different program to run my models. I used this program > to export the coefficient estimates as well as the varcov matrix and > then imported these into Stata. As such, these are the two pieces of > information I currently have: b1 (the coefficient vector) and varcov1 > (the varcov matrix). > > I specified the varcov and the b1 so that price is my first variable > and the ASC (capturing the utility of choosing the status quo option) > is the last. I did this because the help file specifies that the bid > variable should be the first variable specified after wtpcikr and the > constant the last. > > Now, if I type > wtpcikr price_c1 miles5_c1 miles8_c1 miles12_c1 asc_c1, bmat(b1) vmat(varcov1) > > Then I get the error message: > last estimates not found > r(301); > > Can anyone help me with what I am doing wrong? When I look through > your help file, the one thing I see I don't have is the mymean(.) > option. Is this required? If so, how would I obtain this? The data at > hand is from a choice experiment and price_c1 etc. represent the > coefficients on the price variable and miles5_c1 represents the > utility associated with the a level of 5 miles. I'm not sure what > would go in the mymean () vector. > > Any assistance would be very much appreciated! > Have a lovely day. > Sanja > * > * 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/ Dear Sanja, You are right that -wtpcikr- allows you to bring your estimation results into Stata for constructing Krinsky and Robb confidence intervals. However, you need to trick -wtpcikr- in order to have it do what you want. Since you are bringing in your estimation results, in addition to specifying the -bmat()- and -vmat()- options, you need to specify the -mymean()- option to provide the mean values of your variables in a matrix. Here is where the trick comes in. Prior to using -wtpcikr-, you have to estimate in Stata a model with the same distributional assumption you used to get your coefficient estimates. For instance, if you used a logistic distributional assumption to get your estimates, then you need to estimate a simple logit model in Stata. It really does not matter what variables go into the estimation as long as it is a logit model. That step will not only prevent Stata from issuing the r(301) error message but also point -wtpcikr- to the right formula to use for the computations. If you know the mean values of your variables, you can place them into a matrix as follows: . matrix meanval=[val1, val2, val3, val4] If you don't, you can use Stata's -tabstat- command to calculate them and place them into a matrix as follows: . tabstat miles5_c1 miles8_c1 miles12_c1 asc_c1, save . matrix meanval = r(StatTotal) You are now ready to proceed. . logit anyvar anyvar . wtpcikr price_c1 miles5_c1 miles8_c1 miles12_c1 asc_c1, bmat(b1) vmat(varcov1) mymean(meanval) Hope this helps. Best regards, Wilner Rice University 6100 Main Street, Houston, Texas 77005 Email: pwjeanty@rice.edu * * 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/