If you've never seen the Minitab 13 survival/reliability analysis tools
available I strongly recommend you do so (via the downloadable demo at
www.minitab.com).
Both nonparametric and parametric (Normal, Lognormal, Smallest Extreme
Value, Weibull, Exponential, Logistic, Loglogistic). The nonparametric
tools include Kaplan-Meier and actuarial methods. The parametric tools
will all right and abitrary (interval, left, and right) censoring. It can
include weight variable, user defined censor variable (you can use any
value to ID censored observations or identify a time censor or failure
censor value). You can output a full array of graphs, most with
confidence intervals, including probability plots, hazard plots, survival
plots. You can also use a "by" variable to do multiple analyses with a
few clicks.
They are powerful and VERY easy to use - well worth looking at
I've been using reliability software for years (SAS, Reliasoft's
Weibull++, an internal GM SAS module, etc) and Minitab's is right up
there.
I will admit it does have two shortfalls though - no 3-parameter Weibull
and it doesn't automatically output F(t) plots - it only does 1-F(t)
plots (survival). I'm told that should be resolved in version 14.
By the way, it also does regression with life data and accelerated life
testing (with up to two acceleration factors).
I don't work there -- I just like the software lots
Vincent Vinh-Hung wrote:
> I've received email, the performance problem with Survival analysis
> was fixed in JMP version 4.0.2
> a patch exists to move from 4.0.0 to 4.0.2
> John Sall benchmarked 6 groups 162,000 rows,
> version 4.0.2 ran all by groups in 7 seconds total.
> Previously when I ran JMP 3.2, the total time (without time
> spent manipulating the mouse) would have been 18 seconds.
>
> I think this is worth mentionning that these benchmarks are
> doubly remarkable for both JMP 3.2 and 4.
> Firstly is the capability of running very large database,
> and secondly the speed is very impressive - since running
> a Kaplan-Meier (KM) survival analysis implies sorting all the records
> by time. The calculation basis of KM is simple, but implementing
> KM efficiently is something I daren't even imagine the difficulties.
>
> To further comments on Ken's question asking comparison of JMP 3
> and JMP 4. The two coexist on the same system, JMP 4 installation
> didn't interfere with 3.2, files directly readable from one to
> the other. Core system looks the same. Handling data,
> import-export, much improved. Commands are easier, By_group mode
> now immediately available while previously required to run
> first Summary_table. I believe the major and the most exciting
> difference is script possibility in JMP 4, I would expect
> the perspective of libraries of procedures to extend capabilities.
>
> (Comment to Vincent Vinh-Hung who was replying to Ken's
> request for comment from users of JMP 4.0 comparing it to JMP 3
> and maybe Minitab 13).
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
