PDF File of CMAJ paper "categorisation not a bad thing"

[Martin Holt]

Hi,

 

A couple of days ago I posted a comment that a case could be made for first performing categorisation before doing just the continuous analysis (as is always said to be required), as the data might have an unusual shape (eg BMI is parabolic).

 

I've been asked to supply a copy of the CMAJ paper that makes this case. So I've scanned it in as a PDF document, and if anyone else wants a copy please contact me offlist on [hidden email] and I'll duly send one.

 

Best Wishes,

 

Martin Holt
Medical Statistician

[Swank, Paul R]

However, I think it should be pointed out that this is the exception rather than the rule. It should not be seen as a justification for categorizing contiuous data in general, nut only in certain circumstances.

 

Dr. Paul R. Swank,

Professor

Children's Learning Institute

University of Texas Health Science Center-Houston

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Martin Holt
Sent: Wednesday, June 15, 2011 11:07 AM
To: [hidden email]
Subject: PDF File of CMAJ paper "categorisation not a bad thing"

 

Hi,

 

A couple of days ago I posted a comment that a case could be made for first performing categorisation before doing just the continuous analysis (as is always said to be required), as the data might have an unusual shape (eg BMI is parabolic).

 

I've been asked to supply a copy of the CMAJ paper that makes this case. So I've scanned it in as a PDF document, and if anyone else wants a copy please contact me offlist on [hidden email] and I'll duly send one.

 

Best Wishes,

 

Martin Holt
Medical Statistician

 

[Dale Glaser]

...........there was a similar article in Psychological Methods that actually showed, in some circumstances, when dichotomization worked as well as continuous variables...see the following:

DeCoster, J., Iselin, A. R., & Gallucci, M.  (2009).  A conceptual and empirical examination of justifications for dichotomization.  Psychological Methods, 14, 349-366.

Dale Glaser, Ph.D.
Principal--Glaser Consulting
Lecturer/Adjunct Faculty--SDSU/USD/Alliant
Past-President, San Diego Chapter of
American Statistical Association
3115 4th Avenue
San Diego, CA 92103
phone: [phone]
fax: [phone]
email: [hidden email]
website: www.glaserconsult.com

--- On Wed, 6/15/11, Martin Holt <[hidden email]> wrote:

From: Martin Holt <[hidden email]>
Subject: PDF File of CMAJ paper "categorisation not a bad thing"
To: [hidden email]
Date: Wednesday, June 15, 2011, 9:06 AM

Hi,

 

A couple of days ago I posted a comment that a case could be made for first performing categorisation before doing just the continuous analysis (as is always said to be required), as the data might have an unusual shape (eg BMI is parabolic).

 

I've been asked to supply a copy of the CMAJ paper that makes this case. So I've scanned it in as a PDF document, and if anyone else wants a copy please contact me offlist on m861holt@... and I'll duly send one.

 

Best Wishes,

 

Martin Holt
Medical Statistician

[Martin Holt-2]

The paper recommends performing an initial *exploratory* analysis using categorisation to identify those very circumstances when categorisation would be the best way forward.

 

If the analyst already has that information then, yes, it would be expected that the categorical approach would be the exception rather than the rule.

 

But if not, the exploratory analysis would identify some instances when the categorical data approach would be optimal. You'd only know by first checking it out.

 

So does it come down to saving effort....most of the time you'll be OK....which (a) doesn't seem professional to me, and (b) is not the argument usually given for only doing continuous data analysis.

 

I'm aware that  this position goes against the recommendations of the majority (well, asks for an initial exploratory analysis), and that's why I disseminated the reference...it is very persuasive.

 

Best Regards,

 

Martin Holt

Medical Statistician

 

 

 

 

---

From: "Swank, Paul R" <[hidden email]>
To: "[hidden email]" <[hidden email]>; "[hidden email]" <[hidden email]>
Sent: Wed, 15 June, 2011 17:56:07
Subject: RE: PDF File of CMAJ paper "categorisation not a bad thing"

However, I think it should be pointed out that this is the exception rather than the rule. It should not be seen as a justification for categorizing contiuous data in general, nut only in certain circumstances.

 

Dr. Paul R. Swank,

Professor

Children's Learning Institute

University of Texas Health Science Center-Houston

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Martin Holt
Sent: Wednesday, June 15, 2011 11:07 AM
To: [hidden email]
Subject: PDF File of CMAJ paper "categorisation not a bad thing"

 

Hi,

 

A couple of days ago I posted a comment that a case could be made for first performing categorisation before doing just the continuous analysis (as is always said to be required), as the data might have an unusual shape (eg BMI is parabolic).

 

I've been asked to supply a copy of the CMAJ paper that makes this case. So I've scanned it in as a PDF document, and if anyone else wants a copy please contact me offlist on [hidden email] and I'll duly send one.

 

Best Wishes,

 

Martin Holt
Medical Statistician

 

[Swank, Paul R]

I always suggest exploratory analyses. One way to examine data for correlational approaches is by looking at simple scattergrams to see if the data violate the assumption of linearity. There is no excuse for blindly analyzing data without understanding the nature of that data. So by all means, check it out!

 

Dr. Paul R. Swank,

Professor

Children's Learning Institute

University of Texas Health Science Center-Houston

 

From: Martin Holt [mailto:[hidden email]]
Sent: Thursday, June 16, 2011 12:38 PM
To: Swank, Paul R; [hidden email]
Subject: Re: PDF File of CMAJ paper "categorisation not a bad thing"

 

The paper recommends performing an initial *exploratory* analysis using categorisation to identify those very circumstances when categorisation would be the best way forward.

 

If the analyst already has that information then, yes, it would be expected that the categorical approach would be the exception rather than the rule.

 

But if not, the exploratory analysis would identify some instances when the categorical data approach would be optimal. You'd only know by first checking it out.

 

So does it come down to saving effort....most of the time you'll be OK....which (a) doesn't seem professional to me, and (b) is not the argument usually given for only doing continuous data analysis.

 

I'm aware that  this position goes against the recommendations of the majority (well, asks for an initial exploratory analysis), and that's why I disseminated the reference...it is very persuasive.

 

Best Regards,

 

Martin Holt

Medical Statistician

 

 

 

 

 

---

From: "Swank, Paul R" <[hidden email]>
To: "[hidden email]" <[hidden email]>; "[hidden email]" <[hidden email]>
Sent: Wed, 15 June, 2011 17:56:07
Subject: RE: PDF File of CMAJ paper "categorisation not a bad thing"

However, I think it should be pointed out that this is the exception rather than the rule. It should not be seen as a justification for categorizing contiuous data in general, nut only in certain circumstances.

 

Dr. Paul R. Swank,

Professor

Children's Learning Institute

University of Texas Health Science Center-Houston

 

From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Martin Holt
Sent: Wednesday, June 15, 2011 11:07 AM
To: [hidden email]
Subject: PDF File of CMAJ paper "categorisation not a bad thing"

 

Hi,

 

A couple of days ago I posted a comment that a case could be made for first performing categorisation before doing just the continuous analysis (as is always said to be required), as the data might have an unusual shape (eg BMI is parabolic).

 

I've been asked to supply a copy of the CMAJ paper that makes this case. So I've scanned it in as a PDF document, and if anyone else wants a copy please contact me offlist on [hidden email] and I'll duly send one.

 

Best Wishes,

 

Martin Holt
Medical Statistician