Unfortunately I cannot send the data. However, after scrutinizing the
data I found that the SPSS file given to my had automatically divided
by 30.5 instead of 30 when creating the time_variable (month) like I
had done in stata. I get equivalent results between SPSS and stata
using stsum now. Thanks a lot for the information provided, Theresa.
Regards,
M
On Wed, Oct 7, 2009 at 5:44 PM, Theresa Boswell, StataCorp LP
<[email protected]> wrote:
> Moleps Islon <[email protected]> obtained different median survival times from
> the commands -stsum- and -stdescribe-.
>
>>>What is the difference in median time at risk per subject and median
>>>survival time? Which one to report?
>
> From the manual entry for -stdescribe-, we have the following statement:
>
> "stdescribe produces a summary of the st data in a computer or data-based
> sense rather than in an analytical or statistical sense."
>
> The values obtained from -stdescribe- are the same as simply summarizing the
> time variable with the -summarize- command. Here is an example showing that the
> median time from -stdescribe- matches the median obtained from -summarize-.
>
> . webuse page2, clear
> . stdescribe
>
> failure _d: dead
> analysis time _t: time
>
> |-------------- per subject --------------|
> Category total mean min median max
> ------------------------------------------------------------------------------
> no. of subjects 40
> no. of records 40 1 1 1 1
>
> (first) entry time 0 0 0 0
> (final) exit time 227.95 142 231 344
>
> subjects with gap 0
> time on gap if gap 0
> time at risk 9118 227.95 142 231 344
>
> failures 36 .9 0 1 1
> ------------------------------------------------------------------------------
>
> . summarize time, detail
>
> time
> -------------------------------------------------------------
> Percentiles Smallest
> 1% 142 142
> 5% 149.5 143
> 10% 163.5 156 Obs 40
> 25% 201 163 Sum of Wgt. 40
>
> 50% 231 Mean 227.95
> Largest Std. Dev. 46.37691
> 75% 245 296
> 90% 296 304 Variance 2150.818
> 95% 313.5 323 Skewness .3677894
> 99% 344 344 Kurtosis 3.047023
>
> . /* stdescribe matches the median from summarize: 231 */
>
> The command -stsum- reports the median survival time based on analytical
> methods that take into account censoring, delayed entry, and gaps in history.
> This can be reproduced by estimating the survival function and finding the
> time corresponding to median survival. Here is an example:
>
> . sts generate s = s
>
> . summarize s, detail
>
> S(t+0)
> -------------------------------------------------------------
> Percentiles Smallest
> 1% .0296852 .0296852
> 5% .0445277 .0296852
> 10% .0890555 .0593703 Obs 40
> 25% .282009 .0890555 Sum of Wgt. 40
>
> 50% .5127436 Mean .5022545
> Largest Std. Dev. .2874161
> 75% .75 .9
> 90% .8875 .925 Variance .082608
> 95% .9375 .95 Skewness -.0635524
> 99% .975 .975 Kurtosis 1.788548
>
> . sort s
>
> . list time s if s<= `r(p50)' & s>= `r(p50)'-.1
>
> +------------------+
> | time s |
> |------------------|
> 19. | 232 .48575712 |
> 20. | 232 .48575712 |
> +------------------+
>
> . /*stsum matches the value of time where the median survival time occurs :
> 232*/
>
> Thus, the median reported survival time is not the same for the commands
> -stsum- and -stdescribe-. The values from -stsum- are probably what you
> want to use.
>
> As for Moleps' question regarding results obtained from SPSS, we were not able
> to reproduce a difference in reported medians. We ask that Moleps send his
> dataset and commands to our technical support staff at [email protected]
> so that we can look into this issue further.
>
> --Theresa
> [email protected]
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/statalist/faq
> * http://www.ats.ucla.edu/stat/stata/
>
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
