[R] glm.fit and predict.glm: error ' no terms component'

[Thomas Lumley]

On Thu, 30 Sep 2004, Christoph Lehmann wrote:

> many thanks I did it the following way, based on Thomas' suggestion

There was a reason why I didn't call the function predict.glm.fit: it 
isn't a method for predict. You will be calling it directly, rather than 
via predict as you should for a method.

 	-thomas

> predict.glm.fit<-function(glmfit, newmatrix){
>   newmatrix<-cbind(1,newmatrix)
>   coef <- rbind(1, as.matrix(glmfit$coef))
>   eta <- as.matrix(newmatrix) %*% as.matrix(coef)
>   exp(eta)/(1 + exp(eta))
> }
>
>
> cheers
>
> christoph
>
>
>
>
>
>
>
> Thomas Lumley wrote:
>> On Wed, 29 Sep 2004, Christoph Lehmann wrote:
>> 
>>> Hi
>>> 
>>> when I fit a glm by
>>> 
>>>     glm.fit(x,y,family = binomial())
>>>     and then try to use the object for prediction of newdata by:
>>> 
>>>     predict.glm(object, newdata)
>>> 
>>> I get the error:
>>> 
>>> Error in terms.default(object) : no terms component
>>> 
>>> I know I can use glm() and a formula, but for my case I prefer 
>>> glm.fit(x,y)...
>> 
>> 
>> Well, you can't use predict.glm that way.  As the function name suggests, 
>> it is a predict method for objects of class "glm", which in your case you 
>> do not have.
>> 
>> There are two reasons why it won't work.  For type="terms" the formula is 
>> needed to identify terms, and for any type of prediction the formula is 
>> needed to convert the data frame newdata into a model matrix.
>> 
>> You would need to write a function where the new data was a model matrix. 
>> If you only need point predictions then
>> 
>> predict_glm_fit<-function(glmfit, newmatrix, addintercept=TRUE){
>>    if (addintercept)
>>     newmatrix<-cbind(1,newmatrix)
>>    eta<-glmfit$coef %*% newmatrix
>>    family$linkinv(eta)
>> }
>> 
>> would work.
>> 
>>     -thomas
>> 
>> ______________________________________________
>> R-help at stat.math.ethz.ch mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide! 
>> http://www.R-project.org/posting-guide.html
>> 
>> 
>
>

Thomas Lumley			Assoc. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle