[R] How does predict.lm work?

[Marc Schwartz]

on 09/09/2008 09:59 AM Williams, Robin wrote:
> Hi, 
>   Please could someone explain how this element of predict.lm works?
>>From the help file 
> `
> newdata	  
> An optional data frame in which to look for variables with which to
> predict. If omitted, the fitted values are used.
> ' 
>   Does this dataframe (newdata) need to have the same variable names as
> was used in the original data frame used to fit the model? 

Yes. Also, see the Note in ?predict.lm:

"Variables are first looked for in newdata and then searched for in the
usual way (which will include the environment of the formula used in the
fit). A warning will be given if the variables found are not of the same
length as those in newdata if it was supplied."


It also says "Variables", not columns.

> Or will R
> just look across consecutive columns of newdata, and apply them to the
> call as appropriate?

No.

>   For example, if I have fitted a model with four variables
> (x1,x2,x3,x4) in my original dataframe, and then have a second dataframe
> which I want to supply to the newdata argument in predict.lm with
> variable names (x5, x6, x7, x8), do I need to change the variable names
> in my newdata dataframe to match those of the original dataframe? 

Yes.

> Or
> will R treat x5 as x1, x6 as x2, etc, when using predict.lm? 
>   I would like to know so that I can design the structure of some
> somewhat larger dataframes in a manner which will make using predict.lm
> straight forward and quick.
> Hope this makes sense.
> Many thanks for any help. 

HTH,

Marc Schwartz