R help - Apr 2011 - saving a regression model to a file

Hello all,

First off, I am using R version 2.13.0 in Ubuntu.

I have read previous posts in the R mailing list on saving models for
later use, and the responses indicate using the R save() function to
save a model and then using load() to load it is one way to go.
However, when I use the save() function to save a model to a file, and
then try to load it using load(), the loaded "model" is just an object
of class Character. Here is the relevant R code:

#######creating the model#########
data <- read.table("../synthetic_data_4.csv", header=TRUE,
sep=",")

data <- data.frame(data)

data$type <- factor(data$type)

pcrmodel <- pcr(formula=count~week+type+total, data=data, x=TRUE, y=TRUE)

#########getting summary of model########

summary(pcrmodel)

Output:

Data: 	X dimension: 318 4
	Y dimension: 318 1
Fit method: svdpc
Number of components considered: 4
TRAINING: % variance explained
       1 comps  2 comps  3 comps  4 comps
X        99.57   100.00   100.00   100.00
count    13.70    16.01    16.17    16.54


########saving the model#########

save(pcrmodel, file='model.rda')

########loading the model#########

model <- load('model.rda')

########getting summary of model after loading###########

summary(model)

Output:

 Length     Class      Mode
        1 character character

##################################

What am I doing wrong? Or is this not the correct way to save a model
to be used later? Thanks,

Alison

On Apr 14, 2011, at 12:45 PM, Alison Callahan wrote:
> Hello all,
> 
> First off, I am using R version 2.13.0 in Ubuntu.
> 
> I have read previous posts in the R mailing list on saving models for
> later use, and the responses indicate using the R save() function to
> save a model and then using load() to load it is one way to go.
> However, when I use the save() function to save a model to a file, and
> then try to load it using load(), the loaded "model" is just an
object
> of class Character. Here is the relevant R code:
> 
> #######creating the model#########
> data <- read.table("../synthetic_data_4.csv", header=TRUE,
sep=",")
> 
> data <- data.frame(data)
> 
> data$type <- factor(data$type)
> 
> pcrmodel <- pcr(formula=count~week+type+total, data=data, x=TRUE,
y=TRUE)
> 
> #########getting summary of model########
> 
> summary(pcrmodel)
> 
> Output:
> 
> Data: 	X dimension: 318 4
> 	Y dimension: 318 1
> Fit method: svdpc
> Number of components considered: 4
> TRAINING: % variance explained
>       1 comps  2 comps  3 comps  4 comps
> X        99.57   100.00   100.00   100.00
> count    13.70    16.01    16.17    16.54
> 
> 
> ########saving the model#########
> 
> save(pcrmodel, file='model.rda')
> 
> ########loading the model#########
> 
> model <- load('model.rda')
> 

Your error is above, in that you are assigning the result of load() to a new
object and presuming that the new object contains 'pcrmodel'.

If you look at ?load, you will see that the function returns a character vector
of the names of the objects contained in the file you are loading.

Thus, 'model' is a character vector as show below, in this case, likely
containing a single entry of "pcrmodel". Type:

 model

at the R prompt....


If you use:

  summary(pcrmodel)

after the load() call, you should get your normal model summary output. When
using load(), it restores the saved objects in the file to the current
workspace. If you use:

  ls()

you should see 'pcrmodel' listed, along with anything else in your
current workspace.

HTH,

Marc Schwartz


> ########getting summary of model after loading###########
> 
> summary(model)
> 
> Output:
> 
> Length     Class      Mode
>        1 character character
> 
> ##################################
> 
> What am I doing wrong? Or is this not the correct way to save a model
> to be used later? Thanks,
> 
> Alison