R biglm predict searching for dependent variable - r

Question

I'm using the biglm package to run a regression on a data set. The regression runs fine using the following code:
chunkStart <- seq(1,150000000,1000000)
chunkEnd <- seq(1000000,151000000,1000000)
ff <- price ~ factor(Var1) + factor(Var2)
#for(i in 1:length(chunkStart)){
for(i in 1:5){
startRow <- chunkStart[i]
endRow <- chunkEnd[i]
curchunk <- data.frame( price=x[startRow:endRow,1]
,Var1=factor( x[startRow:endRow,6], levels=1:3), Var2= factor( x[startRow:endRow,7], levels=1:3 ) )
if(i == 1){
a <- biglm(ff,curchunk )
}
if(i != 1){
a <- update(a,curchunk )
}
rm(curchunk )
gc()
print(paste(i, " | ",startRow ," | ",endRow ," | ", sep=""))
flush.console()
}
> summary(a)
Large data regression model: biglm(ff, curchunk)
Sample size = 5000000
Coef (95% CI) SE p
(Intercept) 0.0457 0.0454 0.0461 2e-04 0
factor(Var1)2 0.0189 0.0184 0.0194 2e-04 0
factor(Var1)3 0.0148 0.0142 0.0155 3e-04 0
factor(Var2)2 -0.0331 -0.0335 -0.0326 2e-04 0
factor(Var2)3 -0.0417 -0.0426 -0.0408 4e-04 0
The problems come when I try to predict using the biglm object, 'a'.
> df1 <- data.frame(y[1:1000,])
> pred1 <- predict(a, df1)
Error in eval(expr, envir, enclos) : object 'price' not found
Why is the predict function looking for the price/ dependent variable? Any suggestions?
EDIT:
> head(df1)
Var1 Var2
1 3 3
2 3 1
3 3 2
4 2 1
5 2 2
6 1 1
> str(df1)
'data.frame': 1000 obs. of 2 variables:
$ Var1: Factor w/ 3 levels "1","2","3": 3 3 3 2 2 1 2 1 2 1 ...
$ Var2: Factor w/ 3 levels "1","2","3": 3 1 2 1 2 1 1 1 2 1 ...
> pred1 <- predict(a, df1)
Error in eval(expr, envir, enclos) : object 'price' not found

Answer 1

The reason it is looking for the dependent variable is that the predict method uses a call to model.frame from the stats package, and that function requires all the variables to be present in the new data. This is indicated on the model.frame help page without explanation for the motivation behind it.
All you actually need to do about this is create a variable in your new data that has the same name as the dependent variable, then fill it with zeroes (or any non-missing value). So it should work if you run this:
df1$price <- 0
pred1 <- predict(a, df1)