R help - Oct 2002 - automatic chi-square grouping in R

I'm doing some chi-square tests, and I recall some arbitrary rule that says
each band must have at least 5 events in order for the test to be meaningful. Is
there some way to do the banding automagically in R ? For instance, in the
following survdiff, I'm trying to see if ADL affects survival. But when
ADL=3,5 and 6, the number observed is too little. Anyway for me to tell R how to
group them ? Like "R, combine ADL=5 and ADL=6, and redo the test" ?

-----------

Call:
survdiff(formula = Surv(days, status) ~ adl, data = nu)

       N Observed Expected (O-E)^2/E (O-E)^2/V
adl=0 92        6     8.74   0.86134   1.17556
adl=1 38        5     3.41   0.74346   0.83435
adl=2 60        9    10.56   0.22975   0.39159
adl=3 44        4     5.22   0.28487   0.33978
adl=4 27        6     2.32   5.83153   6.30818
adl=5 31        3     3.12   0.00456   0.00506
adl=6 16        2     1.63   0.08385   0.08835

 Chisq= 8.2  on 6 degrees of freedom, p= 0.226 

-------

On a related note, is it possible to tell R to group together values, for
instance, if I have age in my data ranging from 30-60, is it possible to tell R
to convert all ages 30-35 into 32.5, all age from 36-40 into 37.5 ... etc ? I
mean I can always do this in Excel before I feed the data into R, but it seems R
must be able to do something like this. I just don't know where to begin
looking in the manual for something like this ...

Thanks so much guys,

Roger


======   Legal Disclaimer   =====
As you may know, emails sent via Internet can easily be altered or manipulated
by third persons. For this reason, we do not assume any responsibility for
changes made to this message after it was sent.
Furthermore please note that the information (including any attachments) in this
email is strictly confidential and solely intended for the addressee named
above. If you are neither the intended recipient nor an employee or agent
responsible for delivering this message to the intended recipient, we hereby
notify you that any form of unauthorized use, publication, reproduction, copying
or disclosure of this emails contents is not permitted.
If you have received this communication in error, please notify the sender and
delete the email from your computer system.


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._

Hi,
there are possibilities to group, regroup, and transform data in R easily.
For example,

adl2<-adl
adl2[adl2==5 | adl2==6]<-7 #or whatever you want
then
survdiff(formula = Surv(days, status) ~ adl2, data = nu)

To make this automagically, I would use something along these simple lines:

t1<-table(adl)
#then order it (descending)
t2<-t1[rev(order(t1))]
t3<-t2[t2>=5] #or another threshold
remainder<-sum(t2[t2<5])
names(remainder)<-"Blahblah"
mynewadlgrouping<-rbind(t3,remainder) # I think it is correct, if not, then
use cbind() or even simpler c()

and use as.factor(mynewadlgrouping) instead of adl.

But joining groups in order to build bigger groups is sometimes
statistically doubtful.
The same for age :

age[age>30 & age<=35]<-32.5
age[age>35 & age<=40]<-37.5
etc.

Or there is a possibility to build factor, for example

age2<-as.factor(floor(age/5))

JS


- - - Original message: - - -
From: owner-r-help at stat.math.ethz.ch
Send: 11.10.2002 12:03:08
To: <r-help at stat.math.ethz.ch>
Subject: [R] automatic chi-square grouping in R

I'm doing some chi-square tests, and I recall some arbitrary rule that says

each band must have at least 5 events in order for the test to be
meaningful.
Is there some way to do the banding automagically in R ? For instance, in
the
following survdiff, I'm trying to see if ADL affects survival. But when
ADL=3,5
and 6, the number observed is too little. Anyway for me to tell R how to
group
them ? Like "R, combine ADL=5 and ADL=6, and redo the test" ?

-----------

Call:
survdiff(formula = Surv(days, status) ~ adl, data = nu)

       N Observed Expected (O-E)^2/E (O-E)^2/V
adl=0 92        6     8.74   0.86134   1.17556
adl=1 38        5     3.41   0.74346   0.83435
adl=2 60        9    10.56   0.22975   0.39159
adl=3 44        4     5.22   0.28487   0.33978
adl=4 27        6     2.32   5.83153   6.30818
adl=5 31        3     3.12   0.00456   0.00506
adl=6 16        2     1.63   0.08385   0.08835

 Chisq= 8.2  on 6 degrees of freedom, p= 0.226

-------

On a related note, is it possible to tell R to group together values, for
instance, if I have age in my data ranging from 30-60, is it possible to
tell R
to convert all ages 30-35 into 32.5, all age from 36-40 into 37.5 ... etc ?
I
mean I can always do this in Excel before I feed the data into R, but it
seems
R must be able to do something like this. I just don't know where to begin
looking in the manual for something like this ...

Thanks so much guys,

Roger


-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.
-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._.
_._._

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at
stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._