Hi all,
I'm trying to run some monte carlo simulation for my roll call data
using the ideal() function, which resides in the pscl package.
However, I'm receiving an error message that I don't understand.
Error in ideal(a, maxiter = 1000, thin = 10, burnin = 50, store.item = TRUE, :
NA/NaN/Inf in foreign function call (arg 13)
my code is simple the following:
+ store.item = TRUE, normalize=T,
+ priors=list(xp=1e-12,xpv=1e-12,bp=1e-12,bpv=1e-12), verbose=T)
ideal: analysis of roll call data via Markov chain Monte Carlo methods.
Subsetting rollcall object a using dropList
Using the following codes to represent roll call votes:
Yea: 1
Nay: 0
Abstentions: NA
Not In Legislature: 9
Ideal Point Estimation
Number of Legislators 30
Number of Items 1172
checking for any user-supplied priors...
checking start values...
will use eigen-decomposition method to get start values for ideal points...done
running 1172 vote-specific probit GLMs
for start values for item/bill parameters
conditional on start values for ideal points...done
using the following start values for ideal points (summary follows):
V1
Min. :-0.7573
1st Qu.:-0.1583
Median : 0.0000
Mean :-0.0342
3rd Qu.: 0.0000
Max. : 0.8160
using the following start values for item parameters (summary follows):
V1 V2
Min. :-131.2397 Min. :-49.27293
1st Qu.: -1.9090 1st Qu.: -1.33518
Median : -1.5156 Median : -0.34876
Mean : 0.5443 Mean : -0.07001
3rd Qu.: 1.6978 3rd Qu.: 1.22064
Max. : 553.5285 Max. : 65.19820
NA's : 573.0000
Starting MCMC Iterations...
Error in ideal(a, maxiter = 1000, thin = 10, burnin = 50, store.item = TRUE, :
NA/NaN/Inf in foreign function call (arg 13)
+
The problem, I think, is that many data matrix has too many NA. Does
anybody have any suggestion?
Thanks in advance,
Antonio.