Ranking & ties

[<Peter M?>]

Hello to everybody?

I've got a Problem:

I've different ordinal variables with 4 categories, which shoukd be ranked.
There is a given Priorisation (say from 1 to 10). So it is no surprise that
there a lot of ties.
But how can I give a ascending numbers to these ties:

e.g.:
v1 ranks WANTED
4 14 5
4 14 15
4 14 16
4 14 17
4 14 18
3 9 9
2 4 4
1 1 1
3 9 10
2 4 5
1 1 2
1 1 3
2 4 6
2 4 7
2 4 8
3 9 11
3 9 12
4 14 19
3 9 13
4 14 20

Using Excel i can do this without problems
*=RANG(B1; B$1:B$20; 1) + ZÄHLENWENN(B$1:B1; B1) - 1

Does someone give me a hint, how do this with SPSS.
Additionaly I've got some missings.

Thank you

Peter Müller

[Marta Garcia-Granero]

Hi Peter

You wrote:

I've different ordinal variables with 4 categories, which shoukd be ranked.
There is a given Priorisation (say from 1 to 10). So it is no surprise that
there a lot of ties.
But how can I give a ascending numbers to these ties:

e.g.:
v1 ranks WANTED
4 14 5
4 14 15
4 14 16
4 14 17
4 14 18
3 9 9
2 4 4
1 1 1
3 9 10
2 4 5
1 1 2
1 1 3
2 4 6
2 4 7
2 4 8
3 9 11
3 9 12
4 14 19
3 9 13
4 14 20

Using Excel i can do this without problems
*=RANG(B1; B$1:B$20; 1) + ZÄHLENWENN(B$1:B1; B1) - 1

Does someone give me a hint, how do this with SPSS.
Additionaly I've got some missings.


A couple of questions:
- I think there is a typo error in your example. Shouldn't first
"wanted"data be 14 instead of 5?
- Second question: what do you plan to do with missing data: ignore
them, rank them as lowest values or hightes values?

Assuming there really is a typo error, and 5 should be 14 in first row,
here is a solution using MATRIX (I was going to use RANK /CONDENSE
procedure, but it is not giving the expected results).

data list list/v1 WANTED (3 F8).
begin data
4 5
4 15
4 16
4 17
4 18
3 9
2 4
1 1
3 10
2 5
1 2
1 3
2 6
2 7
2 8
3 11
3 12
4 19
3 13
4 14 20
end data.


* No missing data *.
CACHE.
EXE.
MATRIX.
GET data /VAR=v1.
COMPUTE Ranked=GRADE(data).
COMPUTE vname={'Ranked'}.
SAVE Ranked /NAMES=vname /OUTFILE='C:\Temp\RankedData.sav'.
END MATRIX.
MATCH FILES /FILE=*
/FILE='C:\Temp\RankedData.sav'.
FORMAT Ranked(F8).
LIST.

* Missing data are ommited *.
CACHE.
EXE.
MATRIX.
GET data /VAR=v1 /MISSING=ACCEPT /SYSMIS=100.
COMPUTE Ranked=GRADE(data).
COMPUTE vname={'Ranked'}.
SAVE Ranked /NAMES=vname /OUTFILE='C:\Temp\RankedData.sav'.
END MATRIX.
MATCH FILES /FILE=*
/FILE='C:\Temp\RankedData.sav'.
IF MISSING(ranks) Ranked=ranks.
FORMAT Ranked(F8).
LIST.

* Missing data are treated as very low answers *.
CACHE.
EXE.
MATRIX.
GET data /VAR=v1 /MISSING=ACCEPT /SYSMIS=0.
COMPUTE Ranked=GRADE(data).
COMPUTE vname={'Ranked'}.
SAVE Ranked /NAMES=vname /OUTFILE='C:\Temp\RankedData.sav'.
END MATRIX.
MATCH FILES /FILE=*
/FILE='C:\Temp\RankedData.sav'.
FORMAT Ranked(F8).
LIST.

* Missing data are treated as very high answers *.
CACHE.
EXE.
MATRIX.
GET data /VAR=v1 /MISSING=ACCEPT /SYSMIS=100.
COMPUTE Ranked=GRADE(data).
COMPUTE vname={'Ranked'}.
SAVE Ranked /NAMES=vname /OUTFILE='C:\Temp\RankedData.sav'.
END MATRIX.
MATCH FILES /FILE=*
/FILE='C:\Temp\RankedData.sav'.
FORMAT Ranked(F8).
LIST.

Regards,
Marta Garcia-Granero

[cbautista]

Hi,

 

I have to do a survival analysis or time-to-event analysis and in my data some variables are fixed (i.e. gender) and in other variables their values change over time (i.e. CD4).

 

I've heard that Cox regression analysis with time-dependent covariates is the appropiate statistical procedure for this kind of analysis. So, I'm looking for examples in SPSS and guidelines about this procedure. 

 

Unfortunately, in my study, I have some missing data for some time-dependent variables. So in these cases, what is the most common solution? imputation methods? or just not to use these variables at all.

 

Thanks in advance for your comments and suggestions,

 

/Christian

 

  

 

[Hector Maletta]

Christian,

If your study is longitudinal, you probably mean that for a certain subject
you have a missing value for covariate k at time t, but you have a valid
value at times t-1 and t+1 (and other times as well. If that is the case,
and the number of such missing values is not too large, I do not see much to
worry: either you interpolate the missing value (assuming the variable in
question shows a more or less regular pattern over time, and you have the
adjacent values for t-1 and t+1) or you leave it blank for that particular
point in time and that particular subject. In the latter option, that
subject will be included in the analysis but the observation of that subject
at time t will not be included.



Of course, if you have a large number of missing points, you may be in
trouble, because they would reduce the overall number of observation points,
possibly in a biased manner.



Hector







  _____

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of
Christian Bautista
Sent: 07 October 2007 15:26
To: [hidden email]
Subject: Survival analysis (time-to-event analysis)
Importance: High



Hi,



I have to do a survival analysis or time-to-event analysis and in my data
some variables are fixed (i.e. gender) and in other variables their values
change over time (i.e. CD4).



I've heard that Cox regression analysis with time-dependent covariates is
the appropiate statistical procedure for this kind of analysis. So, I'm
looking for examples in SPSS and guidelines about this procedure.



Unfortunately, in my study, I have some missing data for some time-dependent
variables. So in these cases, what is the most common solution? imputation
methods? or just not to use these variables at all.



Thanks in advance for your comments and suggestions,



/Christian