R help - Jul 2010 - help with predict.lda

HI, Dear community,

I am using the linear discriminant analysis to build model and make new
predictions:
> dim(train)  #training data
[1] 1272   22
> dim(valid)  # validation data
[1] 140  22


lda.fit <- lda(out ~ ., data=train, na.action="na.omit", CV=TRUE) #
model
fitting of linear discriminant analysis on training data
> predict(lda.fit, valid)   # make prediction on validation data
Error in UseMethod("predict") :
  no applicable method for 'predict' applied to an object of class
"list"

Can anyone help with this?

Thanks so much!

-- 
Sincerely,
Changbin
--

	[[alternative HTML version deleted]]

On 2010-07-03 21:33, Changbin Du wrote:
> HI, Dear community,
>
> I am using the linear discriminant analysis to build model and make new
> predictions:
>
>> dim(train)  #training data
> [1] 1272   22
>> dim(valid)  # validation data
> [1] 140  22
>
>
> lda.fit<- lda(out ~ ., data=train, na.action="na.omit",
CV=TRUE) # model
> fitting of linear discriminant analysis on training data
>
>> predict(lda.fit, valid)   # make prediction on validation data
> Error in UseMethod("predict") :
>    no applicable method for 'predict' applied to an object of class
"list"
>
> Can anyone help with this?
Suggestions:

0. lda() is not a function in the base installation of R;
    I'll assume that you mean MASS::lda.

1. ask yourself what you expect from setting CV = TRUE.

2. *carefully* (re)read the help pages for lda (especially
    the Value section) and for predict.lda (the 'object'
    definition).

3. use CV = FALSE

4. whenever problems arise, *first* use str() on your
    objects to see what you've got.

5. finally, do provide *reproducible* code; here, you
    could have used the example on the help page.

   -Peter Ehlers
>
> Thanks so much!
>

On Jul 3, 2010, at 11:33 PM, Changbin Du wrote:
> HI, Dear community,
>
> I am using the linear discriminant analysis to build model and make  
> new
> predictions:
>
>> dim(train)  #training data
> [1] 1272   22
>> dim(valid)  # validation data
> [1] 140  22
>
>
> lda.fit <- lda(out ~ ., data=train, na.action="na.omit",
CV=TRUE) #
> model
> fitting of linear discriminant analysis on training data
>
>> predict(lda.fit, valid)   # make prediction on validation data
> Error in UseMethod("predict") :
>  no applicable method for 'predict' applied to an object of class  
> "list"
The predict methods are very picky with what can be used as input.  
They need to have a dataframe with exactly the same variable names as  
were used in the rhs of the formula given to the fitting function. In  
this case that would be all columns in "train" except $out. Reasoning
under the assumption that "valid" looks just like "train" at
the
moment, I think it's possible that you need to remove the "$out"  
column from the "valid" object since it was not on the rhs. Of course
it is also possible that "valid" is even more poorly constructed than
I think, so any further questions should adhere to Peter's admonitions  
as well as including str(train) and str(valid).

>
> Can anyone help with this?
>
-- 

David Winsemius, MD
West Hartford, CT