[Limdep Nlogit List] random effects logistic model
Noel Roy
noelroy at mun.ca
Tue Mar 10 03:05:43 EST 2009
Your lagged regressor model did not converge (status=3), so the
results for this model are meaningless.
The question you have to address is why the lagged regressor model
did not converge. It could be because of the variable substitution
(ROS[-1] for ROS), or it could be because 1996 is dropped. I would
estimate the original model with 1996 dropped to see if this model
converges with 1996 dropped.
At 03:31 PM 08/03/2009, you wrote:
>Hallo,
>
>I'm running a random effects logistic regression. My dependent variable
>is adummy (new_inn) indicating if a firm has or has not introduced an
>innovation in the reference year. Among the regressors ros is a
>profitability ratio. Other regressors are time invariant (inizio,
>north_ea north_we centre).
>
>When the model is estimated with contemporaneous regressors, rho is
>important and quite "big", thus signalling that the contribution of the
>random effect to the total variance is significant (results shown
>below).
>
>--> sample;all$
>--> LOGIT; lhs=new_inn; RHS= one,add,inizio, dredbv,
>INNSET,NORTH_EA,north_we...
> PANEL;PDS=8;PERIOD=time; RANDOM EFFECTS$ ? dredbv, red_add_,
> red_imim,in...
> +--------------------------------------------------+
> | Panel Data Binomial Logit Model |
> | Number of individuals = 2591 |
> | Number of periods = 8 |
> | Random effects model for panel data. |
> +--------------------------------------------------+
>Normal exit from iterations. Exit status=0.
>
>+---------------------------------------------+
>| Logit Model for Panel Data |
>| Maximum Likelihood Estimates |
>| Model estimated: Mar 08, 2009 at 06:51:26PM.|
>| Dependent variable NEW_INN |
>| Weighting variable None |
>| Number of observations 20728 |
>| Iterations completed 40 |
>| Log likelihood function -7360.178 |
>| Hosmer-Lemeshow chi-squared = ********* |
>| P-value= .00000 with deg.fr. = 8 |
>| Random Effects Logit Model for Panel Data |
>| Variance decomposition: Rho = .630189 |
>| Var[e] = pi-squared/3 = 3.289868 |
>| Var[u] = Var[e]*Rho/(1-Rho) = 5.606223 |
>+---------------------------------------------+
>+---------+--------------+----------------+--------+---------+
>|Variable | Coefficient | Standard Error |b/St.Er.|P[|Z|>z] |
>+---------+--------------+----------------+--------+---------+
> Characteristics in numerator of Prob[Y = 1]
> Constant -11.3248235 .49514852 -22.872 .0000
> ADD .00021605 .460196D-04 4.695 .0000
> INIZIO -.01952131 .00308193 -6.334 .0000
> DREDBV .25955100 .05556371 4.671 .0000
> INNSET .18523351 .00482228 38.412 .0000
> NORTH_EA 2.80927316 .21278967 13.202 .0000
> NORTH_WE 2.38145592 .21016688 11.331 .0000
> CENTRE 1.76615375 .21813424 8.097 .0000
> ROS .01751809 .00406118 4.314 .0000
> IMAT .325059D-05 .120828D-05 2.690 .0071
> OFFAT .19390741 .01492164 12.995 .0000
> LEVERAGE -.00073968 .00029694 -2.491 .0127
> Q_IMMMAT -.00022810 .00107902 -.211 .8326
> Rho .63018950 .01070766 58.854 .0000
>
>
>
>when I introduce lagged regressors (for example past profitability
>l1.ros) , rho becomes small and the Likelihood-ratio test of rho=0 is
>significant (results shown below)
>
>--> sample;all$
>--> reject;time=1996$
>--> LOGIT; lhs=new_inn; RHS= one,add,inizio, dredbv,
>INNSET,NORTH_EA,north_we...
> PANEL;PDS=7;PERIOD=time; RANDOM EFFECTS$ ? dredbv, red_add_,
> red_imim,in...
> +--------------------------------------------------+
> | Panel Data Binomial Logit Model |
> | Number of individuals = 2591 |
> | Number of periods = 7 |
> | Random effects model for panel data. |
> +--------------------------------------------------+
>Line search does not improve fn. Exit iterations. Status=3
> Error 806: Line search does not improve fn. Exit iterations. Status=3
>Function= .12571290575D+05, at entry, .61305847660D+04 at exit
>
>+---------------------------------------------+
>| Logit Model for Panel Data |
>| Maximum Likelihood Estimates |
>| Model estimated: Mar 08, 2009 at 06:44:26PM.|
>| Dependent variable NEW_INN |
>| Weighting variable None |
>| Number of observations 18137 |
>| Iterations completed 28 |
>| Log likelihood function -6130.585 |
>| Hosmer-Lemeshow chi-squared = 54.46803 |
>| P-value= .00000 with deg.fr. = 8 |
>| Random Effects Logit Model for Panel Data |
>| Variance decomposition: Rho = .000000 |
>| Var[e] = pi-squared/3 = 3.289868 |
>| Var[u] = Var[e]*Rho/(1-Rho) = .000000 |
>+---------------------------------------------+
>+---------+--------------+----------------+--------+---------+
>|Variable | Coefficient | Standard Error |b/St.Er.|P[|Z|>z] |
>+---------+--------------+----------------+--------+---------+
> Characteristics in numerator of Prob[Y = 1]
> Constant -5.37119662 .25054257 -21.438 .0000
> ADD .870916D-04 .162709D-04 5.353 .0000
> INIZIO -.00723346 .00114506 -6.317 .0000
> DREDBV .21222777 .03454865 6.143 .0000
> INNSET .06772354 .00291373 23.243 .0000
> NORTH_EA 1.29464549 .11161528 11.599 .0000
> NORTH_WE 1.19623940 .11050248 10.825 .0000
> CENTRE .88253976 .11292736 7.815 .0000
> ROS[-1] .01995731 .00250428 7.969 .0000
> IMAT .189168D-05 .631310D-06 2.996 .0027
> OFFAT .08680035 .01040113 8.345 .0000
> LEVERAGE -.00057893 .00016112 -3.593 .0003
> Q_IMMMAT .00073268 .00073521 .997 .3190
> Rho .259782D-15 399458.387 .000 1.0000
>
>
>Is there someone who can help me to interpret these results?
>thank you
>Eleonora Bartoloni
>
>_______________________________________________
>Limdep site list
>Limdep at limdep.itls.usyd.edu.au
>http://limdep.itls.usyd.edu.au
-----
Dr. Noel Roy
Professor and Head
Department of Economics
Memorial University of Newfoundland
St. John's, Newfoundland, Canada A1C 5S7
tel: (709) 737-8245
fax: (709) 737-2094
WWW: http://www.ucs.mun.ca/~noelroy
More information about the Limdep
mailing list