lm3<-glmer.nb(Number_scallops~Stratum+len_scale+
    offset(offset_term)+
    (1|StationID),data=length_data2)

#use ddply to create prediction dataset
library(plyr)
MyData <- ddply(length_data2,
                .(Stratum,offset_term,StationID),
                summarize,
                len_scale = seq(min(len_scale), max(len_scale),length = 25))

#saves scale parameters to convert scaled variable back
scaleList <- list(scale = attr(length_data2$len_scale, "scaled:scale"),
    center = attr(length_data2$len_scale, "scaled:center"))

#prediction
MyData$newdata<-round(predict(lm3,MyData,type="response"),0)

#convert len_scale back to originalscale
MyData$length_o<-round(MyData$len_scale*scaleList$scale + `scaleList$center,0)`