R: predict.lm() not recognizing an object

user1751333 picture user1751333 · Oct 16, 2012 · Viewed 9k times · Source
> reg.len <- lm(chao1.ave ~ lg.std.len, b.div) # b.div is my data frame imported from a CSV file
> reg.len

Call:
lm(formula = chao1.ave ~ lg.std.len, data = b.div)

Coefficients:
(Intercept)   lg.std.len  
      282.4       -115.7  

> newx <- seq(0.6, 1.4, 0.01)
> prd.len <- predict(reg.len, newdata=data.frame(x=newx), interval="confidence", level=0.90, type="response")
Error in eval(expr, envir, enclos) : object 'lg.std.len' not found

I've tried doing the lm like this: lm(b.div$chao1.ave ~ b.div$lg.std.len), but then, predict() gives a warnings that the newdata and variables are different lengths. So, I tried the way above, and now predict() gives an error saying it doesn't recognize the object. How to fix, please?

Answer

Brandon Bertelsen picture Brandon Bertelsen · Oct 16, 2012

Predict expects newdata to have the same column names (to match the formula in reg.len). You're changing it to "x" in your newdata specification, which isn't part of the formula.

dat <- data.frame(y=rnorm(50),lg.std.len=sample(10:15,50,replace=TRUE))
reg.len <- lm(y ~ lg.std.len,data=dat)

newx <- seq(0.6, 1.4, 0.01)
prd.len <- predict(reg.len, newdata=data.frame(lg.std.len=newx),
                   interval="confidence", level=0.90, type="response")

The key part is newdata=data.frame(lg.std.len=newx)