Hi, Who has experience calculate elasticity and partial effect manually for random parameter negative binomial? I have tried many times but didn't get the same value with software. I am using Nlogit 5 software to run my model.
Best Regards,
Rusdi
|Nonrandom parameters
Constant| -2.27660***
B1| .86718***
B2| .06254**
|Dispersion parameter for NegBin distribution
ScalParm| .45631***
Variables Mean
X1 0.078268
X2 0.441176
Partial Effect
--------+--------------------------------------------------------------------
X1| .09792***
X2| .00706*
Hi Rusdi, have you used formulas, presented in Washington et al.'s Statistical and Econometric Methods for Transportation Data Analysis, Second Edition? (chapter 11.2)
Hi Jiri,
Yes. I've been using that formula (11.7) but also get the different result. Maybe I miss something there.
In the book there is a worked example 11.2 - you can test whether you come to the same results.
You can calculate the elasticities and partial effects using the mean values for the variables, or you can calculate them for all observations and take the mean values. This will give you two different estimates. With random parameters, you will need to use the ": Parameters" option in the model code and use each observation to get the same value as the software.
For random parameter you can not calculate marginal effect manually because when the parameter is random each observation have it is own beta,, but the example you showed seemed to be fixed parameters in a random parameter model in that case for x1 marginal effect= beta1* EXP (constant+beta1 X1+beta2X2),, note use mean of X1 and X2 you should get the same results reported by the software and try the same for the second fixed parameter ( go for Washington et al. (2011)) for more details and Anastapolas and Mannering 2009.
Thanks Jwan,
I just realised that I already dropped random parameter in my example.
Anyways thanks for your answer.
Regards
Rusdi
Oh in that case the example that I show you is for fixed parameters model you can not apply this to random parameters model,
- Badges
- Science topic
- Similar topics
- Geoscience
- Asia
- India