RE: st: OBS xtlogit/xtreg

[Nerissa Fanoembi <nerissa_f@hotmail.com>]

I carried out a xtlogit, depvar indepvar fe estimation and the 
estimation output indicates a note that “n groups are dropped because of
 all positive or all negative outcomes”. This means that there is not 
within individual variation in the outcome variable for n individuals. I
 am wondering if there is a command that gives me the number of 
individuals that do not have within individual variation in a variable, 
or as a matter of fact in the dependent variable. The issue here is that
 before the “xtlogit, depvar indepvar, fe” estimation I was busy with 
carrying out “xtreg, depvar indepvar, fe” estimation and the estimation 
output of xtreg does not tell me that n individuals do not have within 
individual variation. Hence I coincidentally discovered after I have run
 the xtlogit estimation that n individuals do not have within individual
 variation. In principle I suppose I could create a variable using a 
code like "by ID: gen newvar = sum(depvar != depvar[_n-1])” which could 
show if values of depvar differ across time within each individual and 
then I could count the cases that this does not happen. But still, I am 
wondering if there is a more formal or simpler Stata command that can be
 used to check how many individuals do not have within individual 
variation in a variable. I also would like to ask why the "xtreg depvar indepvar, fe"
would not give a note like "xtlogit depvar indepvar, fe" gives about the individuals that
have no within individual variation. 

> Date: Mon, 11 Nov 2013 18:19:15 -0500
> Subject: Re: st: OBS xtlogit/xtreg
> From: serjradyakin@gmail.com
> To: statalist@hsphsun2.harvard.edu
> 
> Nerissa, check the messages that xtlogit produces. It does drop groups
> under certain conditions, as the following example illustrates:
> *************************************************
> version 12.0
> sysuse bplong, clear
> xtset patient
> replace bp=bp>146
> replace age=age+1 if mod(_n,2)==0
> xtreg bp age sex, fe vce(robust)
> xtlogit bp age sex, fe
> *************************************************
> Hope this helps.
> 
> Best, Sergiy Radyakin
> 
> 
> On Mon, Nov 11, 2013 at 5:35 PM, Alfonso Sanchez-Penalver
> <alfonso.statalist@gmail.com> wrote:
>> Hi Nerissa,
>>
>> I believe that it is joy possible to answer your question with the information you provide. Please describe the response and explanatory variables better. Also you may want to show your exact commands so that we can get a better picture of the situation you are facing.
>>
>> Best regards,
>>
>> Alfonso Sanchez-Penalver
>>
>>> On Nov 11, 2013, at 5:09 PM, Nerissa Fanoembi <nerissa_f@hotmail.com> wrote:
>>>
>>> Hello,
>>>
>>> I research the effect of background characteristics (education, income and gender for example) on risk preferences over time.
>>>
>>> I want to run the model by using xtreg, fe or xtlogit, fe.
>>>
>>> By running xtreg, fe vce(robust) the obs are 1999 and by running the same regression with xtlogit, fe with and without vce(bootstrap) the obs are 800.
>>>
>>> How is it possible that xtlogit has significantly lower observations than xtreg?
>>>
>>> I am looking forward to hearing from you.
>>>
>>> Best Regards,
>>>
>>> Nerissa.
>>> *
>>> *   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/