MINITAB vs. Bland -Altman

[Cliff Blair]

Hello,

 

I've analyzed the data (51 ovservations) from the Bland-Altman article on logrank tests via MINITAB 14. The chi-square statistic for the logrank test was reported as 7.496 while the chi-square statistic reported by B-A is 6.88.

 

(1) Can anyone explain the discrepancy?

(2) Does anyone know of software that computes the statistic as is done in the B-A article?

 

The steps I used were Stat>Reliability/Survival>Dist. Analysis (right censored)>Non. Par. Dist. Analysis.

 

Any help much appreciated.

 

cliff blair

 

 

[Basilio de Bragança Pereira]

Actually Bland-Altman is the Mandel et al more exact test , the logrank of
Peto is an aproximation to it. The exact test sums the several chi-squares
statistics while the log rank sums the obsrved and expected deaths and
calculates the chi squares at the end.
If you use the log rank and acept the hypothesis you should check with the
exact.
Basilio de Braganca Pereira

Cliff Blair Escreveu:

[SR Millis]

I need your help to settle a debate.

 

My colleague is fitting a model in which the dependent variable is an 8-item measure of post-traumatic amnesia for persons with traumatic brain injury. Each item is scored as 0 or 1 (correct/incorrect)--so, subjects' scores can range from 0 to 8. Covariates are entered in the model to examine which are predictive of or associated with performance on this 8-item scale.

 

My colleague used OLS regression. I'm arguing that poisson regression should have been used because the DV is a non-negative count variable.  What do others think?

 

Thanks,

SR Millis 

Scott R Millis, PhD, MEd, ABPP (CN & RP)
Professor & Director of Research
Rehabilitation Institute of Michigan
Wayne State University School of Medicine
261 Mack Blvd
Detroit, MI 48201
Email: smi...@med.wayne.edu
Tel: 313-993-8085
Fax: 313-745-9854

*********************************************************
This electronic message may contain information that is confidential and/or legally privileged. It is intended only for the use of the individual(s) and entity named as recipients in the message. If you are not an intended recipient of this message, please notify the sender immediately and delete the material from any computer. Do not deliver, distribute or copy this message, and do not disclose its contents or take any action in reliance on the information it contains. Thank you.

[Cliff Blair]

Thanks you so much for your reply. Just a couple of poiints of clarification
if I may.

1. Is the statistic calculated by B-A in their logrank article the one you
characterize as more exact? This is the one that sums the observed and
expected deaths and calculates the chi-square at the end.

2. Is the statistic calculated in the B-A article the one due to Peto?

Sorry not to get it but I think the answer to these last queries will clear
things up for me.

cliff

[Basilio de Bragança Pereira]

Dear Cliff
Sorry for the confusion.
Bland -Altman is the log rank which adds the expected death and observed
death and them calculate the qui-square statistics at the end.
Mandel-Haenzel calculates a qui-square at each event and them adds the
several qui-squares.
I found in
Introductory Medical Statistics -1978 -Richard Mould- Inst of Physics
Publishing
a clear exposition of the differences and how to proceed
Probably the minitab is using the Mandel Haenzel.
Bland -Altman is the logrank.
Basilio

Cliff Blair Escreveu:

[Cliff Blair]

Basilio,

Thanks. That does indeed clarify it for me. Thank you for taking the time to
explain.

cliff

[Ronán Conroy]

The items are all different, so even the equal-interval assumption may not be met. I would model this by ordered logistic regression, though negative binomial regression is an alternative.



Ronán