Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
<[email protected]> |

To |
[email protected] |

Subject |
st: Re: st: Syntax of –cmp- for a probit model with ordered probit selection |

Date |
Wed, 13 Nov 2013 16:01:14 +0900 (JST) |

Dear all, I use the dataset "laborsup" in http://www.stata-press.com/data/r12/ to reproduce my problem. The codes are webuse laborsup,clear gen cate_fem_edu=fem_edu replace cate_fem_edu=1 if cate_fem_edu<11 replace cate_fem_edu=2 if cate_fem_edu>=11 & cate_fem_edu<14 replace cate_fem_edu=3 if cate_fem_edu>=14 cmp setup gen ind1=cond(cate_fem_edu==1,$cmp_probit,$cmp_out) gen ind2=cond(cate_fem_edu==2,$cmp_probit,$cmp_out) gen ind3=cond(cate_fem_edu==3,$cmp_probit,$cmp_out) cmp (fem_work=kids other_inc) (fem_work=kids other_inc) (fem_work=kids other_inc) (cate_fem_edu=male_educ), ind(ind1 ind2 ind3 $cmp_oprobit) Here I made female education (fem_edu) become an ordered categorical variable which is valued {1,2,3}. Variable fem_work is a binary variable which shows whether the individual work or not. I estimate the model outcome equation: fem_work= b1*"kids other_inc" + u selection equation: cate_fem_edu= b2*male_educ+ v I want to estimate the outcome equation with the 3 non-overlap subsample determined by the value of cate_fem_edu. The estimation result first shows the probit and ordered probit regression for individual models. Then it shows the following message: Samples for equations 1 and 2 do not overlap. Removing rho_12 from the model. Samples for equations 1 and 3 do not overlap. Removing rho_13 from the model. Samples for equations 2 and 3 do not overlap. Removing rho_23 from the model. Fitting constants-only model for LR test of overall model fit. Equation/parameter /fem_work/ multiply defined r(110); I tried to fix the problem by changing the code into cmp (fem_work1=kids other_inc) (fem_work2=kids other_inc) (fem_work3=kids other_inc) (cate_fem_edu=male_educ), ind(ind1 ind2 ind3 $cmp_oprobit) where fem_work1,2,3 are just the same as fem_work. Then Stata shows Samples for equations 1 and 2 do not overlap. Removing rho_12 from the model. Samples for equations 1 and 3 do not overlap. Removing rho_13 from the model. Samples for equations 2 and 3 do not overlap. Removing rho_23 from the model. Fitting constants-only model for LR test of overall model fit. initial vector: matrix must be dimension 9 r(503); I spent much time to try to figure out what is wrong with my code but still failed. Does any one know what the problem is and how I can correct it? I really appreciate your help. Kuo, Chiu-Wei --- On Tue, 2013/11/12, [email protected] <[email protected]> wrote: > Dear Buchanan, > > Thank you so much for your suggestion. I really appreciate your kindly reply. > > After using -set trace on-, I found out that in order to estimate the model > outcome equation: outcome_var = b1*var_list1 + u (outcome_var is a binary variable. ) > selection equation: selection_var = b2*var_list2 + v (selection_var is an ordered categorical variable valued {1,2,3}) > the appropriate command should be > gen ind1=cond(selection_var ==1, $cmp_probit , $cmp_out) > gen ind2=cond(selection_var ==2, $cmp_probit , $cmp_out) > gen ind3=cond(selection_var ==3, $cmp_probit , $cmp_out) > cmp (outcome_var = var_list1) (outcome_var = var_list1) (outcome_var = var_list1) (selection_var = var_list2), ind(ind1 ind2 ind3 $cmp_oprobit) > so that the outcome equation is estimated with the subsample of each category of selection_var. > > However, Stata returns the error message "Equation/parameter /outcome_var/ multiply defined r(110);" after the execution of this command. > > I also found a post in the past (http://www.stata.com/statalist/archive/2013-03/msg01041.html) in which similar command was used. I am still trying to figure out what is wrong with mine now. I will try to reproduce this problem with some available Stata datasets later. > > Kuo, Chiu-Wei > > --- On Mon, 2013/11/11, William Buchanan <[email protected]> wrote: > > > It may be useful to use -set trace on- to find out exactly where things are breaking down in the code. If you don't want to dig too far into the code interpretation use -set tracedepth 1- before that (it will show only the higher level commands). Have you been able to reproduce this problem with other datasets that other users could use to replicate your issue? > > > > Sent from my iPhone > > > > > On Nov 11, 2013, at 5:54, <[email protected]> wrote: > > > > > > Dear Buchanan, > > > > > > Thank you so much for your reply. > > > I am sorry that I did not explain my question well. > > > > > > The model I am trying to estimate is > > > outcome equation: outcome_var =b1* var_list1 + u > > > selection equation: selection_var = b2*var_list2 + v > > > where outcome_var is a binomial variable, and selection_var is an ordered categorical variable (the categories are labeled from 1 to 7). > > > > > > Thus I use the command > > > cmp (outcome_var = var_list1) (selection_var = var_list2), ind(selection_var*$cmp_probit $cmp_oprobit) > > > which estimates the outcome equation with a probit model and the selection equation with an ordered probit model, but Stata keeps showing the error message “Indicator for outcome_var must only evaluate to integers between 0 and 8. r(198);” > > > > > > The exactly same command works well when I use other binary variables to be the selection variable, so I think that there may be some problem in the syntax > > > ind(selection_var*$cmp_probit $cmp_oprobit) > > > > > > Maybe the syntax should be write in another way when the selection variable is an ordered categorical variable? I try to figure out the correct syntax but failed. > > > > > > > > > > > >> --- On Mon, 2013/11/11, William Buchanan <[email protected]> wrote: > > >> > > >> If your outcome is binary why would you be using an ordered probit? And for clarity, I was asking about the outcome variable in any/all equations. > > >> > > >> Sent from my iPhone > > >> > > >>> On Nov 11, 2013, at 5:12, <[email protected]> wrote: > > >>> > > >>> Dear Buchanan, > > >>> > > >>> Thank you so much for your suggestion. > > >>> I check the data again, and all the data of the outcome variable is valued 0 or 1. > > >>> Do you think there may be any other possible problem? > > >>> > > >>> Kuo, Chiu-Wei > > >>> > > >>>> --- On Mon, 2013/11/11, William Buchanan <[email protected]> wrote: > > >>>> > > >>>> What are the values of your outcome variable? Are any of the values outside of the 0-8 range indicated by the error message? > > >>>> > > >>>> Sent from my iPhone > > >>>> > > >>>>> On Nov 11, 2013, at 4:53, <[email protected]> wrote: > > >>>>> > > >>>>> Dear Tamas, > > >>>>> > > >>>>> Thank you so much for the suggestion. > > >>>>> But the same error message still pops out when I put double quotes on the expression. > > >>>>> (while in the case where the selection variable is a binomial one, the command works well even without double quotes!) > > >>>>> > > >>>>> Do you think there may be any other problem of the syntax? > > >>>>> > > >>>>> Kuo, Chiu-Wei > > >>>>> > > >>>>>> --- On Mon, 2013/11/11, "Bartus Tamás" <[email protected]> wrote: > > >>>>>> > > >>>>>> Dear Kuo, > > >>>>>> > > >>>>>> Instead of > > >>>>>> > > >>>>>> ind(selection_var*$cmp_probit $cmp_oprobit) > > >>>>>> > > >>>>>> try > > >>>>>> > > >>>>>> ind("selection_var*$cmp_probit" $cmp_oprobit) > > >>>>>> > > >>>>>> Expressions appearing in the indicators option must be enlosed in double quotes > > >>>>>> > > >>>>>> Tamas > > >>>>>> > > >>>>>>> On 13/11/11, [email protected] wrote: > > >>>>>>> > > >>>>>>> > > >>>>>>> Dear all, > > >>>>>>> > > >>>>>>> I am trying to estimate the following selection model. > > >>>>>>> outcome equation: outcome_var =b1* var_list1 + u > > >>>>>>> selection equation: selection_var = b2*var_list2 + v > > >>>>>>> where outcome_var is a binomial variable, and selection_var is an ordered categorical variable. > > >>>>>>> > > >>>>>>> I use the command –cmp- (which can be obtained by –ssc cmp-) to estimate the model. My command is > > >>>>>>> cmp (outcome_var = var_list1) (selection_var = var_list2), ind(selection_var*$cmp_probit $cmp_oprobit) > > >>>>>>> However, the error message “Indicator for outcome_var must only evaluate to integers between 0 and 8. r(198);” keeps popping out. > > >>>>>>> > > >>>>>>> I found out that the same error massage appears whenever the selection variable is an ordered categorical variable, while it disappears and the command works normally when the selection variable is a binomial one. > > >>>>>>> > > >>>>>>> In the paper (“Estimating fully observed recursive mixed-process models with cmp”) written by the –cmp- writer, David Roodman, he mentioned that –cmp- is capable of doing the same estimation as –ssm- and –oheckman- do, so I think that –cmp- is the appropriate command for my model. > > >>>>>>> > > >>>>>>> I suppose that there is some problem with the part of the syntax, > > >>>>>>> ind(selection_var*$cmp_probit $cmp_oprobit) > > >>>>>>> I spent much time to figure out what the correct syntax is but still failed. > > >>>>>>> I really appreciate it if anyone can point out the problem of my syntax. > > >>>>>>> Thank you so much. > > >>>>>>> > > >>>>>>> Kuo, Chiu-Wei > > >>>>>>> > > >>>>>>> * > > >>>>>>> * 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/ > > >>>>>> > > >>>>>> -- > > >>>>>> > > >>>>>> Tamás Bartus, PhD > > >>>>>> Associate Professor, Deputy Director > > >>>>>> Institute of Sociology and Social Policy > > >>>>>> Program Director, Doctoral School of Sociology > > >>>>>> Corvinus University, Budapest > > >>>>>> 1093 Budapest, Közraktár utca 4-6. > > >>>>>> Room 424. > > >>>>>> Phone: +36-1-482-7301 > > >>>>>> Fax: +36-1-482-7348 > > >>>>>> Homepage: http://web.uni-corvinus.hu/bartus > > >>>>>> > > >>>>>> * > > >>>>>> * 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/ > > >>> > > >>> * > > >>> * 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/ > > > > * > > * 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/

- Prev by Date:
**Re: st: How to get -project- to log in text instead of smcl.** - Next by Date:
**Re: st: foreach or forvalues to re-structure data** - Previous by thread:
**st: esttab with svy: tabulate** - Next by thread:
**st: estimating treatment effect of a binary endogenous regressor on binary outcome** - Index(es):