|
|  |
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: Re: your mail
If you look at the results headers, you will see that they are expressed
in different forms:
"Log-normal regression -- accelerated failure-time form"
"Weibull regression -- log relative-hazard form"
Therein lies the clue to your answer.
AFT form gives you time until failure whereas log relative hazard gives
you the rate at which failure occurs. Longer times until failure translate
into lower rates of failure. If I remember correctly, the Weibull results
are essentially exp(-b*p) whereas the log-normal are just b if you think
of y=e(xb)u, with y=time until failure.
In short, try the time option for the weibull model.
On Wed, 24 Jan 2007, Benedikt London wrote:
Date: Wed, 24 Jan 2007 18:05:42 +0000
From: Benedikt London <[email protected]>
Reply-To: [email protected]
To: [email protected]
Hi! I am working with panel data using Stata v7. In order to estimate the
probability of a site adopting a certain technology I am using hazard-rate
specifications.
Running the Lognormal regression (streg - dist(lognormal)) I seem to be
getting inverse results to the one's from Weibull (streg - dist(weibull))
(see results below).
Any pointers helping me understand the differences would be much appreciated.
Thanks,
Benedikt.
174 . streg SC_locus_core_MS PC_Complex_Core PC_Complex_Peri_DA
PC_Complex_Peri_EA
SC_complex_change_lbase SC_arc_change SC_comp_change IND_G_Service
IND_G_Manu
f LOG_Emp IT_Intensity IT_Dev LOG_Server LOG_Ntwrk, dist(weibull)
...
Weibull regression -- log relative-hazard form
No. of subjects = 4989 Number of obs =
4989
No. of failures = 1713
Time at risk = 8356
LR chi2(14) =
5724.20
Log likelihood = -669.04683 Prob > chi2 =
0.0000
------------------------------------------------------------------------------
_t | Haz. Ratio Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------------
SC_locus_c~S | 7.67e+08 2.15e+11 0.07 0.942 2.83e-230
2.08e+24
PC_Complex~e | .9599559 .0263458 -1.49 0.136 .9096833
1.013007
PC_Comple~DA | .9993526 .0359207 -0.02 0.986 .9313719
1.072295
PC_Comple~EA | .9137738 .0218846 -3.77 0.000 .8718718
.9576895
SC_com~lbase | 1.054812 .1046327 0.54 0.591 .8684401
1.281181
SC_arc_cha~e | .6622062 .1495885 -1.82 0.068 .4253161
1.031038
SC_comp_ch~e | 1.035305 .0576771 0.62 0.533 .9282131
1.154753
IND_G_Serv~e | .9587473 .0807536 -0.50 0.617 .8128473
1.130835
IND_G_Manuf | .9930614 .0859979 -0.08 0.936 .8380368
1.176763
LOG_Emp | 1.038066 .0395384 0.98 0.327 .9633937
1.118526
IT_Intensity | 2.359489 .7533794 2.69 0.007 1.261919
4.411687
IT_Dev | .3495068 .2044643 -1.80 0.072 .1110442
1.100057
LOG_Server | .9911896 .0342807 -0.26 0.798 .9262273
1.060708
LOG_Ntwrk | .9999756 .0420713 -0.00 1.000 .9208255
1.085929
-------------+----------------------------------------------------------------
/ln_p | 1.154546 .0189167 61.03 0.000 1.11747
1.191622
-------------+----------------------------------------------------------------
p | 3.172583 .0600148 3.05711
3.292418
1/p | .3152006 .0059626 .3037282
.3271063
------------------------------------------------------------------------------
178 . streg SC_locus_core_MS PC_Complex_Core PC_Complex_Peri_DA
PC_Complex_Peri_EA
> SC_complex_change_lbase SC_arc_change SC_comp_change IND_G_Service
IND_G_Manu
> f LOG_Emp IT_Intensity IT_Dev LOG_Server LOG_Ntwrk, dist(lognormal)
failure _d: SC_locus_core_MS
analysis time _t: t
...
Log-normal regression -- accelerated failure-time form
No. of subjects = 4989 Number of obs =
4989
No. of failures = 1713
Time at risk = 8356
LR chi2(14) =
5515.62
Log likelihood = -542.18236 Prob > chi2 =
0.0000
------------------------------------------------------------------------------
_t | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+----------------------------------------------------------------
SC_locus_c~S | -2.784133 32.25093 -0.09 0.931 -65.99479
60.42653
PC_Complex~e | .0127173 .0086475 1.47 0.141 -.0042315
.0296661
PC_Comple~DA | -.0019033 .0114566 -0.17 0.868 -.0243578
.0205512
PC_Comple~EA | .0309895 .0076347 4.06 0.000 .0160258
.0459533
SC_com~lbase | -.0346015 .0338963 -1.02 0.307 -.1010369
.031834
SC_arc_cha~e | .1680083 .0770845 2.18 0.029 .0169254
.3190911
SC_comp_ch~e | -.0167932 .0202866 -0.83 0.408 -.0565542
.0229678
IND_G_Serv~e | .010267 .0278289 0.37 0.712 -.0442766
.0648106
IND_G_Manuf | -.0014168 .0289121 -0.05 0.961 -.0580834
.0552498
LOG_Emp | -.0113104 .0126465 -0.89 0.371 -.0360972
.0134764
IT_Intensity | -.2025714 .0977681 -2.07 0.038 -.3941934
-.0109494
IT_Dev | .2683782 .2040441 1.32 0.188 -.1315409
.6682973
LOG_Server | -.0004663 .0116539 -0.04 0.968 -.0233074
.0223749
LOG_Ntwrk | -.0021339 .014077 -0.15 0.880 -.0297244
.0254566
_cons | 2.876312 32.25103 0.09 0.929 -60.33455
66.08717
-------------+----------------------------------------------------------------
/ln_sig | -1.102429 .0170846 -64.53 0.000 -1.135914
-1.068944
-------------+----------------------------------------------------------------
sigma | .3320636 .0056732 .3211284
.3433711
------------------------------------------------------------------------------
_________________________________________________________________
Get Hotmail, News, Sport and Entertainment from MSN on your mobile.
http://www.msn.txt4content.com/
*
* 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/
- Fred
Note updated contact info (from August 01, 2005 to August 01, 2007):
Frederick J. Boehmke
Associate Professor of Political Science
University of Iowa (on leave)
[email protected]
_-------------------------------------------------------------_
< Robert Wood Johnson Foundation Research Fellow >
< University of Michigan Office: M2232 (SPH II) >
< School of Public Health Phone: 734 764 2089 >
< Health Management and Policy Fax: 734 936 9813 >
< 109 Observatory, Room M2232 [email protected] >
<_ Ann Arbor, MI 48109-2029 _>
-------------------------------------------------------------
http://myweb.uiowa.edu/fboehmke
*
* 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/
