R help - Apr 2006 - lambda, uncertainty coefficient (& Somers D)

Dear colleagues in R,

Has anybody implemented the

1) (Goodman & Kruskal) lambda

or the

2) (Thiel's) uncertainty coefficient

statistics (in the asymmetric and symmetric forms), or is anyone aware 
that they might reside in some package? A search in the R archives 
does indicate that they are (somehow) part of the CoCo package, but I 
would rather not start learning how to transform my data into 
CoCo-format in order to access these functions, regardless of whether 
the CoCo versions are actually intended for calculating the actual 
statistic or for some other package internal purposes, as may 
sometimes be the case.

Furthermore, it appears to me that the 'somers2' function in the Hmisc 
package applies Somers' D only to 2x2 and not larger tables. Am I 
mistaken, or does there exist somewhere else an implementation of the 
Somers' D statistic for the more general RxC tables? This was queried 
in 1999, but no response seemed then to be forthcoming.

Thanks and regards,

 	-Antti Arppe
-- 
=====================================================================Antti Arppe
- Master of Science (Engineering)
Researcher & doctoral student (Linguistics)
E-mail: antti.arppe at helsinki.fi
WWW: http://www.ling.helsinki.fi/~aarppe

Antti Arppe wrote:
> Dear colleagues in R,
> 
> Has anybody implemented the
> 
> 1) (Goodman & Kruskal) lambda
> 
> or the
> 
> 2) (Thiel's) uncertainty coefficient
> 
> statistics (in the asymmetric and symmetric forms), or is anyone aware 
> that they might reside in some package? A search in the R archives 
> does indicate that they are (somehow) part of the CoCo package, but I 
> would rather not start learning how to transform my data into 
> CoCo-format in order to access these functions, regardless of whether 
> the CoCo versions are actually intended for calculating the actual 
> statistic or for some other package internal purposes, as may 
> sometimes be the case.
> 
> Furthermore, it appears to me that the 'somers2' function in the
Hmisc
> package applies Somers' D only to 2x2 and not larger tables. Am I 
> mistaken, or does there exist somewhere else an implementation of the 
> Somers' D statistic for the more general RxC tables? This was queried 
> in 1999, but no response seemed then to be forthcoming.
Somers' D is for a 2-category response variable.  X may have any number 
of categories as long as they are ordered.

Frank Harrell
> 
> Thanks and regards,
> 
>  	-Antti Arppe

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

On Tue, 2006-04-18 at 16:40 +0300, Antti Arppe wrote:
> Dear colleagues in R,
> 
> Has anybody implemented the
> 
> 1) (Goodman & Kruskal) lambda
> 
> or the
> 
> 2) (Thiel's) uncertainty coefficient
> 
> statistics (in the asymmetric and symmetric forms), or is anyone aware 
> that they might reside in some package? A search in the R archives 
> does indicate that they are (somehow) part of the CoCo package, but I 
> would rather not start learning how to transform my data into 
> CoCo-format in order to access these functions, regardless of whether 
> the CoCo versions are actually intended for calculating the actual 
> statistic or for some other package internal purposes, as may 
> sometimes be the case.
> 
> Furthermore, it appears to me that the 'somers2' function in the
Hmisc
> package applies Somers' D only to 2x2 and not larger tables. Am I 
> mistaken, or does there exist somewhere else an implementation of the 
> Somers' D statistic for the more general RxC tables? This was queried 
> in 1999, but no response seemed then to be forthcoming.
> 
> Thanks and regards,
> 
>  	-Antti Arppe

Antti,

Under a separate e-mail to you, I have sent a text file containing R
code for the above and other related association measures.

If anyone else is interested in these, please let me know and I will be
more than happy to forward them on. I did not want to consume a lot of
bandwidth here.

HTH,

Marc Schwartz

Dear Marc Schwartz,

I would also like to request the R code for doing this nominal measure of
association analyses (if you dont mind..)

Your kindly helps are really appreciated.

Thank you in advance.

Nora M.

--
View this message in context:
http://r.789695.n4.nabble.com/lambda-uncertainty-coefficient-Somers-D-tp798112p3562722.html
Sent from the R help mailing list archive at Nabble.com.

On May 31, 2011, at 6:18 AM, Nora M wrote:
> Dear Marc Schwartz,
>
> I would also like to request the R code for doing this nominal  
> measure of
> association analyses (if you dont mind..)
You should learn to search:

RSiteSearch is a function that you can use at the R command line

RSiteSearch("Somers D"")

Or you can install the sos package

require(sos)
findFn("Somers D")

Which does result in more hits, but not quite as focussed or  
informative results.
>
> Your kindly helps are really appreciated.
>
> Thank you in advance.
>
> Nora M.
David Winsemius, MD
West Hartford, CT

On May 31, 2011, at 8:58 AM, David Winsemius wrote:
> 
> On May 31, 2011, at 6:18 AM, Nora M wrote:
> 
>> Dear Marc Schwartz,
>> 
>> I would also like to request the R code for doing this nominal measure
of
>> association analyses (if you dont mind..)
> 
> You should learn to search:
> 
> RSiteSearch is a function that you can use at the R command line
> 
> RSiteSearch("Somers D"")
> 
> Or you can install the sos package
> 
> require(sos)
> findFn("Somers D")
> 
> Which does result in more hits, but not quite as focussed or informative
results.
> 
>> 
>> Your kindly helps are really appreciated.
>> 
>> Thank you in advance.
>> 
>> Nora M.
> 
> David Winsemius, MD
> West Hartford, CT

Nora, if you are seeking to communicate with a specific person, it is best to
directly e-mail that person rather than posting to a list of several thousand
people. Notably, because there is a reasonable chance that the specific person
of interest is not going to see the e-mail and therefore not respond to your
query.

I will e-mail you the file of interest separately.

Regards,

Marc Schwartz

On Oct 24, 2011, at 11:18 PM, hongwm wrote:
> Dear Marc,
> 
> I would also like to request the R code for doing this nominal measure of
> association analyses, thanks in advance. Your kindly helps are really
> appreciated. 
> 
> My email is hongwm at mun.ca
> 
> Have a great day,
> Hong
Hong, if you are seeking to communicate with a specific person, it is best to
directly e-mail that person rather than posting to a list of several thousand
people. Notably, because there is a reasonable chance that the specific person
of interest is not going to see the e-mail and therefore not respond to your
query.

I have e-mailed you the file of interest separately.

Regards,

Marc Schwartz