Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: Duration models (STREG ): Need help

From   "Oleksandr Nikolsko-Rzehvskyy" <>
Subject   st: Duration models (STREG ): Need help
Date   Wed, 12 Feb 2003 13:37:50 +0200

Dear all,

If you can, please, help me. Iím going to investigate banks' failure (in Ukraine) using duration analysis (with time varying covariates). 
What data I have:
Quartile data on banks' population, with various characteristics of each bank. I know when a bank enters (ENTER=1) my sample, and when it leaves (FAIL=1) it. I defined the following vars:
1. TIME, (0..15). Describes a period of time when an observation takes place.
2. ENTER, 1 if a banks enters the sample this period. However, I didnít set (maybe, mistakenly) ENTER=1 for the first period (TIME=1), and set ENTER=1 only when a bank enters a sample in TIME>1
3. FAIL, 1 if it went bankrupt that period
4. bank's ID and other characteristics

What have I done:
i. declared the data as duration:
	stset TIME, failure(FAIL) id(ID) enter(ENTER)
ii. and estimated this model in this way:
	streg  X Y Z, dist(LOGLOGISTIC)

1. Unfortunately, I have no help on STATA (only help-file with syntax description), and I am not sure that Iím estimating what I would like to. If it possible, could you please tell whether I am right, and if not - suggest another way to investigate the data I have (but using duration models). 
2. Also, I hope that STREG reports coefficients which influence hazard ratio (not duration time), and "+" just means that that variable influence positively the prob of failure. Is this true? And what's on earth this "Gamma" (and "1/ln_Gamma") at the bottom of the table? Does a negative sign before Gamma change the meaning of signs of reported coefficients? (if you know the exact functional form of hazard this model (loglogistic) estimates, with that gammas, etc, I will be really grateful to get it)

Thank you in advance, and sorry for stupid questions

PS. I am not sure that my Eng is good enough, thatís why I add a tentative structure of my set:

id		time	enter	fail         some_var
bank_1	1	0	0	0              1
bank_2	3	0	0	0
bank_3	5	0	0	0
bank_1	3	1	0	1
bank_2	4	1	0	0
bank_3	2	1	0	1
bank_4	1	1	1	0

*   For searches and help try:

© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index