Dear all,
I am having trouble estimating a simple (one nest) nested logit model with purely individual-specific data using STATA 10.
I am relating takeover probabilities to firms' accounting data (for example firm leverage, short lev). The dependent variable (atd) takes on 1 if no takeover, 2 if friendly and 3 if hostile.
With this setup I get:
. nlogit atd lev || lev1: || atd:, case(gvkey)
note: 216 cases dropped because they have only one alternative
depvar atd must be a 0-1 variable indicating which alternatives are chosen
Even though redundant, I then created a data set with three rows per firm (each value of the dependent) and a "decision" variable indicating which value of the dependent is correct. For the explanatory variables I simply replicate each value three times. I get the following error:
. nlogitgen lev1 = ATD (takeover: 2|3, notakeover: 1)
new variable lev1 is generated with 2 groups
label list lb_lev1
lb_lev1:
1 takeover
2 notakeover
. nlogit Decision LEV || lev1: || ATD:, case(gvkey)
note: 17 cases (105 obs) dropped due to no positive outcome or multiple positive outcomes per case
variable LEV is not alternative-specific: it has no within-case variability
r(459);
Additionally I tried setting up the data set with the STATA function "fillin" recommended in some postings. This also creates three observations. In contrast to my data set above I do not get three equivalent observations per firm, but two observations with missing value for the options not chosen.
. fillin coname atd
. browse
. rename _fillin chosen
. replace chosen = 1 - chosen
(604 real changes made)
. nlogitgen lev1 = atd (takeover: 2|3, notakeover: 1)
new variable lev1 is generated with 2 groups
label list lb_lev1
lb_lev1:
1 takeover
2 notakeover
. nlogit chosen lev || lev1: || atd:, case(gvkey)
note: 216 cases dropped because they have only one alternative
note: 8 cases (17 obs) dropped due to no positive outcome or multiple positive outcomes per case
no cases remain after removing invalid observations
r(2000);
Theoretically I do not see why my estimation should not work. I'm running out of ideas, so any help would be appreciated.
Thanks a lot, Benjamin
--
Psssst! Schon das coole Video vom GMX MultiMessenger gesehen?
Der Eine f�r Alle: http://www.gmx.net/de/go/messenger03
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
