Only a few suggestion
2. IMHO -xttobit- is still better than -xtreg- even if you don't have censored observation
4. I have Stata 9.2, in which you are warned *against* adding dummis for the panel indicators
Sorry for not being so much useful,
Nicola
P.S. I'll NOT receive/read any email but the Digest.
At 02.33 13/05/2009 -0400, somsupa Nopprach wrote:
>Dear Statalisters,
>
> my dependent variable (y)is an index , whose value range from -2 to 2.
>my data is panel data and i set industry and country as my id. and year as t. i already set tsset.
>
>due to the limit of my y variable, i select to use xttobit. and i interpreted the effect of my Xs on y by using mfx compute, predict(e(-2,2))
>
> I have five questions
>1. is it proper to find the effect of Xs on y by using mfx compute, predict(e(-2,2))
>2. because there is no left and right cencored obs (zero observation) , I wonder that is it still necessary for me to use xttobit. some ofã??my equations have only two censor observationsã??but most of them are zero.
>
>3. my y variable left-side skewed distribution. i have read that it should be normal distribution to allow Tobit efficient. so , how i should manage with this problem. I tried vce option, but when i use vce option the error below is shown.
>"xtintreg_d2 failed to compute scores required by the robust option
>r(504);"
>what i should do to deal with this error?
>4. in my model i also add country dummy and year variable to detrend, is it still appropriate to use random-effects model?
>5. regarding xtreg command, if i use re option(random model), which one is the R2 of the random model in xtreg?
>
>Could you suggest which model i should to use and help me solves the problems above.
>
>thank you so much in advance,
>somsupa
*
* 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/
