Tags: analysis, butas, interested, matlab, programming, quitelimited, statistical, statistics, survival, toolbox

Survival Analysis in Matlab

On Programmer » Matlab

2,768 words with 2 Comments; publish: Tue, 06 May 2008 20:58:00 GMT; (20062.50, « »)

I am interested in doing all my statistical analysis in Matlab, but

as far as I can tell, the statistics toolbox seems to be quite

limited. Does anyone know whether Matlab (or other users) has

tools/functions for "Survival Analysis". Survival Analysis is also

called Reliability/event history/failure-time analysis in various

disciplines.

Thanks,

Mike

All Comments

Leave a comment...

  • 2 Comments
    • >I am interested in doing all my statistical analysis in Matlab, but

      > as far as I can tell, the statistics toolbox seems to be quite

      > limited. Does anyone know whether Matlab (or other users) has

      > tools/functions for "Survival Analysis". Survival Analysis is also

      > called Reliability/event history/failure-time analysis in various

      > disciplines.

      Mike, the Statistics Toolbox has a few things:

      - ecdf to compute the emprical cumulative distribution function, using the

      Kaplan-Meier estimate with censored data (also see ecdfhist)

      - dfittool, normfit, wblfit, and others for fitting parametric distributions

      to censored data (similar option in ksdensity for a nonparametric smoothed

      fit)

      - coxphfit for fitting Cox proportional hazards regression models that

      relate survival times to predictor variables

      The toolbox also has some demos related to survival analysis.

      -- Tom

      #1; Tue, 06 May 2008 20:59:00 GMT
    • Tom Lane wrote:

      >

      > but

      > also

      various

      > Mike, the Statistics Toolbox has a few things:

      > - ecdf to compute the emprical cumulative distribution function,

      > using the

      > Kaplan-Meier estimate with censored data (also see ecdfhist)

      > - dfittool, normfit, wblfit, and others for fitting parametric

      > distributions

      > to censored data (similar option in ksdensity for a nonparametric

      > smoothed

      > fit)

      > - coxphfit for fitting Cox proportional hazards regression models

      > that

      > relate survival times to predictor variables

      > The toolbox also has some demos related to survival analysis.

      > -- Tom

      >

      >

      These are great if you only have data that is censored from above.

      But for us in (re)insurance and I'm sure those in the life sciences,

      we also need the ability to have data that is truncated from below

      too. In (re)insurance, this equates to having a deductible and a

      policy limit. These are used all the time when we have to fit to a

      distribution and calculate the survival curve via Kap-Meier.

      Since the ecdf, dfitool, mle, etc.. don't support truncating from

      below these great tools are for the most part useless to us.

      I would love to see these in a future version of the Stat's toolbox.

      Stephen

      #2; Tue, 06 May 2008 21:00:00 GMT