Hi
I have tried to estimate race winning probabilities with mlogit in R. I have
different amount of contestors in the races and mlogit has a bug so that in
those situations the mlogit does not work. So i tried to add dummy contestors to
the race so that every race has an equal amount of contestors.
The problem is that the dummies infect the estimates.
In the attached excel file i have an examble of the situation. The first sheet
has no dummies and the second sheet has the same information but dummies are
added. When i run mlogit as followin
library(RODBC)
library(mlogit)
library(foreign)
z<-odbcConnectExcel("D:\\Testi2.xls")
y<-sqlFetch(z,"eidummyj?")
Mallidata=mlogit.data(y,choice="Voittaja",shape="long",id.var="P?iv?jal?ht?",alt.var="Kilpailunumero")
summary(mlogit(Voittaja ~ Onkokaikkikeng?tpois + Onkoosakengist?jalassa+
OurChoicedummy +MvaiN+OvaiR+Log-1 , data=Mallidata))
Coefficients :
Estimate Std. Error t-value Pr(>|t|)
Onkokaikkikeng?tpois -16.7722 25041.3517 -0.0007 0.9995
Onkoosakengist?jalassa 38.9587 8371.6220 0.0047 0.9963
OurChoicedummy 39.5846 26049.3329 0.0015 0.9988
MvaiN -41.4435 10733.2019 -0.0039 0.9969
OvaiR 58.0948 11534.2609 0.0050 0.9960
Log -6.5563 4870.5160 -0.0013 0.9989
Log-Likelihood: -1.5546e-07
And the dummies run
Coefficients :
Estimate Std. Error t-value Pr(>|t|)
Onkokaikkikeng?tpois -15.4053 31857.0476 -0.0005 0.9996
Onkoosakengist?jalassa 40.0112 17493.0896 0.0023 0.9982
OurChoicedummy 39.3606 34019.6023 0.0012 0.9991
MvaiN -40.4852 13794.8400 -0.0029 0.9977
OvaiR 57.0165 14268.6747 0.0040 0.9968
Log -5.5429 9380.3919 -0.0006 0.9995
Log-Likelihood: -1.7162e-07
So the likelihoods are different and estimates also.
How should i add the dummies or is there some other way of doing this...?
Ville