s-news
[Top] [All Lists]

Re: factor levels

To: "Frank E Harrell Jr" <f.harrell@vanderbilt.edu>, <s-news@lists.biostat.wustl.edu>
Subject: Re: factor levels
From: "natalie nicholls (IAH-C)" <natalie.nicholls@bbsrc.ac.uk>
Date: Fri, 25 Feb 2005 15:33:12 -0000
Thread-index: AcUbTaHIqHEQ5a9JTrWxSNis5+HN7gAAN70z
Thread-topic: [S] factor levels
sorry, repeated for clarifcation:

> test<-coxph(Surv(cull,status) ~ YoB,na.action=na.omit,SDF2)

> test
Call:
coxph(formula = Surv(cull, status) ~ YoB, data = SDF2, na.action = na.omit)


       coef exp(coef) se(coef)       z    p 
YoB1  3.263    26.136   24.427  0.1336 0.89
YoB2  1.004     2.729    8.149  0.1232 0.90
YoB3  0.227     1.255    4.080  0.0556 0.96
YoB4  0.136     1.146    2.452  0.0556 0.96
YoB5 -0.772     0.462    3.989 -0.1935 0.85
YoB6  0.188     1.207    1.279  0.1471 0.88
YoB7  0.204     1.226    0.959  0.2127 0.83
YoB8  0.340     1.405    0.745  0.4568 0.65

Likelihood ratio test=18  on 8 df, p=0.0208  n=122 (1 observations deleted due 
to missing values)

YoB :  "1991", "1992", "1993", "1994", "1995", "1996", "1997", "1998", "1999"

------------------------------------------------------------------
original question:

> so is Splus comparing each year to 1991 (so YoB1 is 1992), or to 1999 (so 
> YoB1 is 1991), or just to each previous year in the list (so YoB1 is 1992, 
> and is compared to 1991, with YoB2 being compared to YoB1 anad so on)?

Reply from Prof. Frank E Harrell Jr

Please read the posting guide; show your code.  It would also be good to 
use intuitive category names.

You might look at the Design library which has a wrapper function for 
coxph called cph.  The output makes the coding apparent.  cph uses the 
first level as the reference cell.  For many things you don't need to 
know that (e.g., contrast, summary, plot, nomogram, latex functions in 
Design).

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University


Thanks
Natalie

<Prev in Thread] Current Thread [Next in Thread>