non-parametric, non-terminating time-dependent variables

[J Scelza]

I have a data set in which we are observing employment data. We are using
non-parametric date because the employment status is recorded at several
different times, the actual dates of which vary from client to client (see
example data below). Clients also enter and exit the study at different
times. But our research question does not involve terminating data, for
which a survival analysis model (Cox, Kaplan-Meier, etc.) would've been useful.

So my next question is, what is the ideal analysis for our research
question? We want to observe the amount of time between change in employment
status, either for individual clients, a group of clients with similar data,
or both. (We're especially interest in illustrating this with graphs!)

An excerpt of our data looks something like this:

Client ID          Date_of_Observation        Employment_Status

   100                    8/17/05                     1
   100                    3/9/06                      0
   200                    7/5/05                      0
   200                    1/31/06                     0
   300                    11/9/05                     1
   300                    2/8/06                      0
   300                    5/19/06                     1
   400                    1/30/06                     1
   400                    3/7/06                      1
   400                    4/20/06                     0
   400                    6/16/06                     0

Some notes:

--- The employment status is coded as 1 = employed, 0 = unemployed.
--- The length of the study is from July 1, 2005 - June 30, 2006.
--- Not all clients enter nor leave the study at the same time and so, we do
not know what their employment status is prior to their first date of
observation (when that status is then recorded).
--- In the end, we will have clients who's employment status 1) never
changes, 2) changes from one status to another at some point, or 3) changes
from one status to another several times.
--- And, at this point, we are simply observing two variables: time and
employment status. We do not have any covariates in our research question yet.


Any help is greatly appreciated. Thanks.

[ViAnn Beadle]

What determines how many times a client appears in the file? Why are there 4
records for client 400 but only 2 records for client 100 and 3 for client
300.

It sounds like you want the client as a unit of observation here but you
have a basic problem in that a client could have several "episodes of work".
Approaches might be to count the number of episodes per client or count the
total time employed during the period? You might want to so a time series
analysis of sorts to find out the percentages of your clients employed on a
month to month basis for the 12 months. Each of these examples will require
that the data be transformed in some

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of J
Scelza
Sent: Wednesday, September 26, 2007 10:01 AM
To: [hidden email]
Subject: non-parametric, non-terminating time-dependent variables

I have a data set in which we are observing employment data. We are using
non-parametric date because the employment status is recorded at several
different times, the actual dates of which vary from client to client (see
example data below). Clients also enter and exit the study at different
times. But our research question does not involve terminating data, for
which a survival analysis model (Cox, Kaplan-Meier, etc.) would've been
useful.

So my next question is, what is the ideal analysis for our research
question? We want to observe the amount of time between change in employment
status, either for individual clients, a group of clients with similar data,
or both. (We're especially interest in illustrating this with graphs!)

An excerpt of our data looks something like this:

Client ID          Date_of_Observation        Employment_Status

   100                    8/17/05                     1
   100                    3/9/06                      0
   200                    7/5/05                      0
   200                    1/31/06                     0
   300                    11/9/05                     1
   300                    2/8/06                      0
   300                    5/19/06                     1
   400                    1/30/06                     1
   400                    3/7/06                      1
   400                    4/20/06                     0
   400                    6/16/06                     0

Some notes:

--- The employment status is coded as 1 = employed, 0 = unemployed.
--- The length of the study is from July 1, 2005 - June 30, 2006.
--- Not all clients enter nor leave the study at the same time and so, we do
not know what their employment status is prior to their first date of
observation (when that status is then recorded).
--- In the end, we will have clients who's employment status 1) never
changes, 2) changes from one status to another at some point, or 3) changes
from one status to another several times.
--- And, at this point, we are simply observing two variables: time and
employment status. We do not have any covariates in our research question
yet.


Any help is greatly appreciated. Thanks.