Error in estimation of conditional CDF

[AsalPilehvari]

The bandwidth has been selected using "npcdensbw". Now for estimating the conditional CDF, I
execute this code:
Fhat.low <- matrix(predict(npcdist(bws = bw),newdata=mydat.eval.low),neval,neval)

And, I get this error:
Error in (tbw$sfactor[[i]])[tl[[i]]]/dfactor[[i]] : non-numeric argument to binary operator

How can I solve it?

[JKPeck]

Not enough information here to respond. You are getting an error from deep in the np package, but you are apparently running your own R code. Without the details about the estimated model and your code, I can only guess that you have a variable of the wrong type in the input.

[AsalPilehvari]

Here is the code:

#pce00 is a continuous variable of income
#hb.ema is numeric variable related to health which is continuous as well

ethnic: is a factor variable

hdeduc00: is an ordered variable related to the education level

First I run "npcdensbw" to get the bandwidth:

bandwidth selection

mydat$pce00.share <- mydat$pce00/sum(mydat$pce00)
mydat$hb.ema.share <- mydat$hb.ema/sum(mydat$hb.ema)

bw <- npcdensbw(pce00.share+hb.ema.share~
factor(ethnic)+
ordered(hdeduc00),
bwmethod="cv.ml",
cykertype="epanechnikov",
oxkertype="liracine",
uxkertype="liracine",
data=mydat )
save(bw,file="bw.RData")
summary(bw)

Next, using the above estimated bandwidth, the conditional pdf and CDF is estimating using the below code.

ethnic <- "Jawa"
hdeduc00.low <- 2
hdeduc00.high <- 12
neval <- 50
plot.quantile <- 0.90

Generate sequences for the evaluation data that span the range of

observed data.

pce00.seq <- seq(min(mydat$pce00.share),quantile(mydat$pce00.share,plot.quantile),length = neval)
hb.ema.seq <- seq(min(mydat$hb.ema.share),quantile(mydat$hb.ema.share,plot.quantile),length = neval)

Create evaluation data frames for the low and high education groups

mydat.eval.low <- data.frame(expand.grid(pce00.share = pce00.seq,hb.ema.share = hb.ema.seq),ethnic=ethnic,hdeduc00=hdeduc00.low)

Estimate the conditional density and distribution for low education level.

fhat.low <- matrix(predict(npcdens(bws = bw),newdata=mydat.eval.low),neval,neval)
Fhat.low <- matrix(predict(npcdist(bws = bw),newdata=mydat.eval.low),neval,neval)

This is all I execute.
I hope it can give you enough information to address the problem.
Thank you very much.

[JKPeck]

It appears that you are not using the STATS NONPAR REGR procedure at all but running the npcdensbw package directly. I suggest that you direct your question to the maintainer of that package:
Maintainer: Jeffrey S. Racine racinej@mcmaster.ca