Memisc package: imported varibles from SPSS have got wrong measurement

An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120725/57c6f0bb/attachment.pl>

Dear list members,

I have got another problem. I imported an SPSS file with the Memisc  
package
using the following commands:

mz <- spss.system.file("myspssfile.sav")

mz <- subset(mz,select=c(
bsex,balt,xurb,dtaet,kartab,bgeb,boseit,bgeblan,xnuts2,kausb,xerwstat,
asbper,asbhh,ajahr,aquartal,bstaat,xwieoft,gew1,apkz,bpkzm,bpkzv))
The memisc package help file for spss.system.file() (actually labeled  
"importers") says that there is an S4 method for "subset" there does  
not seem to be a separate page describing its behavior or values
Afterwards I checked the measurements of the variables
What does that phrase mean? What code did you actually use?
and they are all
right for most of them (e.g. the variable containing the sex of a  
person is
"nominal" and the variable containing the year is "interval"). For  
two of
the variables the measurement is not o.k, though. They exclusively  
contain
numbers in the SPSS file (e.g. the age of a person) - and no NAs -  
but have
got the measurement "nominal"!
Hard to say. There is no R storage mode that is called "nominal".  
Perhaps some sort of memisc-specific terminology? Or perhaps something  
in your spss dataset... to which you have not provided access? Or the  
default setting for the spss.system.file access method?  I know that  
the Hmisc package's describe() function will report out a variable as  
though it were categorical if there are only 8 or fewer unique values.

After looking at further help pages and trying the help pages code, I  
am guessing that some of my puzzlement might be answered by reading  
the vignettes, but you can do that yourself.

Here's a hackish guess ... try:

?measurement
measurement(mz$age_variable) <- "interval"
# where age_variable is the unstated item in that "select" list
David Winsemius, MD
David's ?measurement
measurement(mz$age_variable) <- "interval"
# where age_variable is the unstated item in that "select" list
is what I use in similar circumstances.

Where it seems to come from is the SPSS users habit of setting value
labels on various categories of user-missing values - so a survey will
commonly have no actual missings (in spss, system-missing) but -9 as
'refused', -8 as 'not contactable' and so on. The importer picks up
the existence of value labels and sets the mode as "nominal" - which
gets transformed into factor in R usage - using base read.spss would
be likely to read these in as factor.

For analysis purposes, these values would be likely to be NA, but it
may be important to record that you were making that change.
On Jul 25, 2012, at 9:48 AM, Marion Wenty wrote:

Dear list members,

I have got another problem. I imported an SPSS file with the Memisc
package
using the following commands:

mz <- spss.system.file("myspssfile.sav")

mz <- subset(mz,select=c(
bsex,balt,xurb,dtaet,kartab,bgeb,boseit,bgeblan,xnuts2,kausb,xerwstat,
asbper,asbhh,ajahr,aquartal,bstaat,xwieoft,gew1,apkz,bpkzm,bpkzv))

The memisc package help file for spss.system.file() (actually labeled
"importers") says that there is an S4 method for "subset" there does not
seem to be a separate page describing its behavior or values

Afterwards I checked the measurements of the variables

What does that phrase mean? What code did you actually use?

and they are all
right for most of them (e.g. the variable containing the sex of a person
is
"nominal" and the variable containing the year is "interval"). For two of
the variables the measurement is not o.k, though. They exclusively contain
numbers in the SPSS file (e.g. the age of a person) - and no NAs - but
have
got the measurement "nominal"!

Hard to say. There is no R storage mode that is called "nominal". Perhaps
some sort of memisc-specific terminology? Or perhaps something in your spss
dataset... to which you have not provided access? Or the default setting for
the spss.system.file access method?  I know that the Hmisc package's
describe() function will report out a variable as though it were categorical
if there are only 8 or fewer unique values.

After looking at further help pages and trying the help pages code, I am
guessing that some of my puzzlement might be answered by reading the
vignettes, but you can do that yourself.

Here's a hackish guess ... try:

?measurement
measurement(mz$age_variable) <- "interval"
# where age_variable is the unstated item in that "select" list

--

David Winsemius, MD

______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
An embedded and charset-unspecified text was scrubbed...
Name: not available
URL: <https://stat.ethz.ch/pipermail/r-help/attachments/20120727/86b46a2e/attachment.pl>