survival analysis/time-dependent variables

From: Juliet Brown (juliet.brown_at_verizon.net)
Date: 09/29/04 

Date: Wed, 29 Sep 2004 17:53:38 +0000 (UTC)

I am in need of statistical syntax help. I am completing my doctoral
dissertation. I have a set of data on which I would like to run a
survival analysis in SPSS 11.0. I have 20 quarters of data for a set
of psychiatric consumers. I would like to run an extended Cox
regression analysis on the data. The reason the model needs to be
extended is that in my covariate set, I have 4 covariates whose values
for each consumer vary over time. It would therefore not make sense
for me to use baseline values for these variables. Presently, each of
the 4 variables is represented by 20 'subvariable' COLUMNS of data
(e.g., the variable"bsi" is represented by "bsi1", "bsi2",... through
"bsi20"), one column/subvariable for each quarter. Thus, each consumer
now has 1 row of data across 80 columns (20 columns for each of the 4
variables). I am in need of help defining via syntax which values for
these 4 variables the analysis should use for each quarter. I would
like to define a new, time-segmented variable for each of the 4
variable sets. Each time-segmented function should define 20 segments
(quarters). The functions would tell the analysis which values to use
for each quarter. E.g., for the variable set "bsi1" through "bsi20",
if "quarter" = 1 then use "bsi1"; if "quarter" = 2 then use "bsi2", an
so on. Here is a sample model with a variable called "c" transformed
into 4 variables, "c1" through "c4". The example therefore defines 4
time segments:

1 <= T_ <=4 , Variable(i) = { C1 C2 C3 C4 }
For 0 T_ <= 1 use C1,
For 1 > T_ <= 2 use C2
For 2 > T_ <= 3 use C3
For 3 > T_ <= 4 use C4

What this implies is an expression of the form:

COMPUTE T_COV_ = ( T_<=1)*C1 + (1>T_<=2)*C2 + (2>T_<=3)*C3 +
(3>T_<=4)*C4 .

For T_ = 1 ( TRUE )*C1 + ( FALSE )*C2 + ( FALSE )*C3 + ( FALSE
)*C4 = C1.
         2 ( FALSE)*C1 + ( TRUE )*C2 + ( FALSE )*C3 + ( FALSE
)*C4 = C2.
         3 ( FALSE)*C1 + ( FALSE )*C2 + ( TRUE )*C3 + ( FALSE
)*C4 = C3.
         4 ( FALSE)*C1 + ( FALSE )*C2 + ( FALSE )*C3 + ( TRUE
)*C4 = C4.

I am in need of someone in the Philadelphia area to help. Feel free to
contact me via email (juliet.brown@verizon.net).