How to do a median split on a variable

[John Reece]

>Hello all:
>
>A simple (I hope) "how to" question:
>
>I need to do a median split on a series of variables before I conduct an
>analysis of those variables. However I haven't been able to figure out how
>to do the split. I'm using SPSS for Windows, version 6.0.
>
>Thanks in advance for any suggestions.
>
>---Ron
>rpo...@csulb.edu

Do you know what the medians are? If not, the EXAMINE procedure will give
them to you. Do you need to create a new variable with the median split
info in it (i.e. 1 = above the median; 2 = below the median)? If so, you
could try:

COMPUTE NEWVAR = 0.
IF (OLDVAR GE X) NEWVAR = 1. Note: NEWVAR = the new median split
IF (OLDVAR LT X) NEWVAR = 2. variable; OLDVAR = the variable you
you are splitting; X = the median.

This assumes that you are going to code values equal to the median as 1s.
If you want to put values on the median in the lower group, then swap GE
for GT and
LT for LE.

The COMPUTE NEWVAR = 0. can be a bit tricky if you have missing values.

Do you want to analyse the values above and below the median separately? Try:

SELECT IF (OLDVAR GE X).
or
SELECT IF (OLDVAR GT X).
depending on what you have done with values equal to the median.

Hope this helps.

**************************************************************************
* John Reece, PhD *
* Department of Psychology & Intellectual Disability Studies *
* Royal Melbourne Institute of Technology *
* Bundoora Campus *
* PO Box 71 Phone: 03-468-2512 *
* Bundoora Victoria 3083 Fax: 03-468-2303 *
* AUSTRALIA Internet/Aarnet: re...@eos.xx.rmit.edu.au *
**************************************************************************

[ron powell]

[steve]

I'm not sure you'll want to thank me for this, but my advice is to
avoid carving up continuous data so you can fit them to an ANOVA scheme
(unless this is a transformation for hopelessly skewed data). There is
a substantial literature on the harm that median splits can do, e.g.,
lower reliability of variable; loss if variable information; type I and
type II errors, depending on the distribution of the data being carved
up. Instead, fit the continuous data to a regression framework, which
is the more flexible parent of the ANOVA.

======================================================================
Steve Owen BITNET: SVO@UCONN
Bureau of Educational Research INTERNET: S...@UCONNVM.UCONN.EDU
U-4, University of Connecticut PH: 203/486-3322
Storrs, CT 06269 FAX: 203/486-0210

[Jonathan Fry]

In article <Pine.NXT.3.90.941005...@tern.csulb.edu> ron powell <rpo...@CSULB.EDU> writes:
>Date: Wed, 5 Oct 1994 21:15:36 -0700
>From: ron powell <rpo...@CSULB.EDU>
>Subject: How to do a median split on a variable

>Hello all:

>---Ron
>rpo...@csulb.edu

You can create median split variables completely automatically using the
dialogs in SPSS for Windows.

From the Transform menu, select Rank Cases. Select the variables to be split,
and, in the Rank Types subdialog, unselect Rank and select NTiles, specifying
two divisions.

I tried that using two variables from the BANK.SAV file, producing this output:

      
-> RANK
-> VARIABLES=age edlevel (A) /NTILES (2) /PRINT=YES
-> /TIES=MEAN .



From New
variable variable Label
-------- -------- -----

AGE NAGE NTILES of AGE
EDLEVEL NEDLEVEL NTILES of EDLEVEL

Because both variable had ties at the median, the new variables had slightly
more 1 values than 2 values.

-------------
Jonathan B. Fry, Developer
SPSS Inc.
j...@spss.com