 |
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
st: issue using starting values in gllamm
Dear Stata list members
I am estimating a 2 level mixed response model using panel data. The
response is dichotomous (0,1) and a function of observed dummy variables,
with direct and indirect effects, and ordered categorical latent variables
(Chapter 6 GLLAMM Manual). Using ?gllamm- (in Stata 10.1) I have run the
following model.
eq id1: op1 op2 op3 yvar
eq id2: op4 op5 op6 op7 yvar
eq id3: op8 op9 op10 op11 yvar
eq f1: male ter nochild
eq f2: male ter nochild
eq f3: male ter nochild
gllamm resp yvar bid1 ter2 male2 nochild2 op2 op3 op5 op6 op7 op9 op10
op11 , nip(8) nrf(3) i(id) eqs(id1 id2 id3) geqs(f1 f2 f3) link(ologit
ologit ologit logit) family(binomial binomial binomial binomial) lv(var)
fv(var) from(ex2) trace adapt
Equations id1-id3 are the measurement models, f1-f3 the ?exposure? models.
The model estimated and appeared to converge (results reported below). To
check the robustness of parameters at different ?nip()-, I retrieved the
parameter matrix (ex2) to use as starting values. However, after using
?from(ex2)- to input the starting values, at iteration 0 it reports:
Updated log likelihood:
. Log-likelihood cannot be computed
finish running on 21 Apr 2009 at 09:42:00
This occurs even if -nip(8)- is used. Any idea why that may be so?
Your help is appreciated.
Regards
Fiona Gibson
PhD Candidate
School of Agricultural and Resource Economics
University of Western Australia
Email: [email protected]
Model results
number of level 1 units = 5688
number of level 2 units = 474
Condition Number = 1334.7088
gllamm model
log likelihood = -5522.7565
----------------------------------------------------------------------------
--
resp | Coef. Std. Err. z P>|z| [95% Conf.
Interval]
-------------+--------------------------------------------------------------
--
resp |
yvar | -.9929185 .2627392 -3.78 0.000 -1.507878
-.4779591
bid1 | -.0087789 .0013141 -6.68 0.000 -.0113545
-.0062032
ter2 | .5641199 .2353094 2.40 0.017 .102922
1.025318
male2 | -.3873411 .2378501 -1.63 0.103 -.8535186
.0788365
nochild2 | .4173029 .2298836 1.82 0.069 -.0332606
.8678665
op2 | .301384 .1795849 1.68 0.093 -.0505958
.6533639
op3 | .4976204 .2281599 2.18 0.029 .0504352
.9448056
op5 | -.3044933 .1789783 -1.70 0.089 -.6552844
.0462978
op6 | -.1403718 .1687842 -0.83 0.406 -.4711828
.1904391
op7 | -.5554956 .1942089 -2.86 0.004 -.9361381
-.1748532
op9 | -.410213 .1346565 -3.05 0.002 -.6741349
-.1462911
op10 | 2.12092 .1470227 14.43 0.000 1.832761
2.409079
op11 | -.5361621 .5208446 -1.03 0.303 -1.556999
.4846744
-------------+--------------------------------------------------------------
--
_cut11 |
_cons | -8.396816 .7351916 -11.42 0.000 -9.837765
-6.955867
-------------+--------------------------------------------------------------
--
_cut12 |
_cons | -4.806453 .6712836 -7.16 0.000 -6.122145
-3.490762
-------------+--------------------------------------------------------------
--
_cut13 |
_cons | .6496311 .6330394 1.03 0.305 -.5911034
1.890366
-------------+--------------------------------------------------------------
--
_cut14 |
_cons | 5.940093 .6770886 8.77 0.000 4.613024
7.267162
-------------+--------------------------------------------------------------
--
_cut21 |
_cons | -6.057203 .602005 -10.06 0.000 -7.237111
-4.877295
-------------+--------------------------------------------------------------
--
_cut22 |
_cons | -2.362765 .5666456 -4.17 0.000 -3.47337
-1.25216
-------------+--------------------------------------------------------------
--
_cut23 |
_cons | 3.128984 .5711984 5.48 0.000 2.009455
4.248512
-------------+--------------------------------------------------------------
--
_cut24 |
_cons | 7.178246 .5999195 11.97 0.000 6.002426
8.354067
-------------+--------------------------------------------------------------
--
_cut31 |
_cons | -4.64904 .3124781 -14.88 0.000 -5.261486
-4.036594
-------------+--------------------------------------------------------------
--
_cut32 |
_cons | -1.25689 .2775795 -4.53 0.000 -1.800936
-.7128441
-------------+--------------------------------------------------------------
--
_cut33 |
_cons | 1.441071 .2777721 5.19 0.000 .8966474
1.985494
-------------+--------------------------------------------------------------
--
_cut34 |
_cons | 5.005855 .3132992 15.98 0.000 4.3918
5.61991
----------------------------------------------------------------------------
--
Variances and covariances of random effects
----------------------------------------------------------------------------
--
***level 2 (id)
var(1): 40.248971 (4.970909)
loadings for random effect 1
op1: 1 (fixed)
op2: 1.0893465 (.03925325)
op3: 1.2308001 (.04589239)
yvar: .10182486 (.03433342)
cov(2,1): 27.819774 (2.9062462) cor(2,1): .77503958
var(2): 32.011399 (3.5252376)
loadings for random effect 2
op4: 1 (fixed)
op5: .90412739 (.03409143)
op6: .98922153 (.03568108)
op7: .81991618 (.03114715)
yvar: .12849134 (.03833614)
cov(3,1): 5.5007943 (.98454576) cor(3,1): .32871944
cov(3,2): 4.2710618 (.83762777) cor(3,2): .28619425
var(3): 6.9573726 (.84022282)
loadings for random effect 3
op8: 1 (fixed)
op9: 1.1164339 (.06203392)
op10: 1.0281634 (.06511839)
op11: -.97757119 (.05905692)
yvar: -.20807531 (.05384573)
Regressions of latent variables on covariates
----------------------------------------------------------------------------
--
random effect 1 has 3 covariates:
male: .78137295 (.63633932)
ter: 1.218253 (.62910364)
nochild: .21101281 (.62815516)
random effect 2 has 3 covariates:
male: 1.7181385 (.56427469)
ter: 1.2200142 (.5511976)
nochild: .44838222 (.55413639)
random effect 3 has 3 covariates:
male: -.22595219 (.26684788)
ter: .48065322 (.26684636)
nochild: .03505078 (.26306724)
----------------------------------------------------------------------------
--
*
* 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/