Date:         Mon, 26 Sep 2005 14:39:10 +0100
Reply-To:     Margaret MacDougall <margmacd@talk21.com>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Margaret MacDougall <margmacd@talk21.com>
Subject:      Clarification on linear trend test
Content-Type: text/plain; charset=iso-8859-1

Dear SPSS users

I would be grateful for a clarification on what test is actually being assumed on selecting the SPSS option to choose 'Linear trend to factor levels' under 'Comparing Factor Levels' when carrying our a Kaplan Meier Analysis. Previously (through reading publications) I had assumed that this was the Chi-Square test for a linear trend which one can equally well implement using the SPSS Crosstabs procedure for running a Chi-Square test of association.  However, the two tests do not deliver the same p-values.  The (albeit naive) reader using the relevant dialogue box options for a Kaplan Meier analysis is led to assume by the way the options are presented that they are performing a log-rank test for a linear trend. Does such a test exist and if so, could someone kindly provide a reference.

On a related note, if the two tests (Chi-Square test for linear trend and 'Log-rank test for linear trend') are distinct, I would be interested to learn which is the superior one in testing for monotonicity of gradient across different severity categories of a factor when predicting mortality.

Many thanks

Margaret


---------------------------------
Yahoo! Messenger  NEW - crystal clear PC to PC calling worldwide with voicemail