Current life table vs Clinical life table

[PETER LUO]

It seems that SAS and SPSS can only produce clinical life table, but
not period life table. For example, I have a data set that contains
customer infomation with a cable company in 1994. Of them some were
disconnected during 1994 after subscribing the service for a period of
time(event), some were still active by the end of 1994(censored). I
was expecting a current life table in which the disconnection rate for
each time interval is determined by dividing the inactive customers by
all customers who had subscribed the service for that long.

years Inactive Active Total Disconnection Rate
0-1 34 123 157 34/157
1-2 15 89 104 15/104
...

Howver, when i use proc lifetest, the procedure will treat the
customers who have been with the service for two years and are still
active as censored cases for the 0-1 years period, thus inflates the
denominator of the 0-1 years. Instead of 34/157, now it's 34/(157+89).

My question is: is there a way to get around this problem? I believe
the current life table method is correct. And since i only have 1994
data, current life table is my only choice. I speculate if i have a
longitudinal data, e.g. all customers who started subscribing the
service in 1980 and their connection status in the following 20 years.
Then i can use clinical life table method and get the same result as
what i got from 1994 period data by using current life table method.

[Richard F Ulrich]

Dear Peter,

It seems to me that you are wanting to see a simple Cross-tabulation,
rather that something that *I* would call a life-table. The statistics
that SPSS or SAS produces will be appropriate to 'life-tables', in the
way that I know them.

So, I do not understand what your problem is... the *rate* of
disconnections for a year certainly includes, as base, *all* the
people who subscribed for that many years OR MORE, even though you
seem to object to that idea.

Rich Ulrich, biostatistician wpi...@pitt.edu
Western Psychiatric Inst. and Clinic Univ. of Pittsburgh
======================start of question
PETER LUO (PL...@DRAFT.DRAFTNET.COM) wrote:
: It seems that SAS and SPSS can only produce clinical life table, but

: not period life table. For example, I have a data set that contains
: customer infomation with a cable company in 1994. Of them some were
: disconnected during 1994 after subscribing the service for a period of
: time(event), some were still active by the end of 1994(censored). I
: was expecting a current life table in which the disconnection rate for
: each time interval is determined by dividing the inactive customers by
: all customers who had subscribed the service for that long.

: years Inactive Active Total Disconnection Rate
: 0-1 34 123 157 34/157
: 1-2 15 89 104 15/104
: ...

: Howver, when i use proc lifetest, the procedure will treat the
: customers who have been with the service for two years and are still
: active as censored cases for the 0-1 years period, thus inflates the
: denominator of the 0-1 years. Instead of 34/157, now it's 34/(157+89).

: My question is: is there a way to get around this problem? I believe

=================================== . . .