I would like to compare two Kenadalls' tau coefficients. Is it possible to test whether one coefficient is significantly higher than the other one ?
Thanks. * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
Why don't you simply construct a 95% confidence interval around the tau
estimate and check if estimates overlap? STATA reports the ASE for tau with the tab y x, all command. HTH, John Antonakis On 04.12.2008 12:34, Marcus Fischer wrote: > I would like to compare two Kenadalls' tau coefficients. Is it possible to test whether one coefficient is significantly higher than the other one ? > Thanks. > > * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
-somersd- by Roger Newson can compute the difference between two
Kendalls Tau coefficients. It is downloadable from SSC. On Dec 4, 2008, at 8:39 AM, JOHN ANTONAKIS wrote: > Why don't you simply construct a 95% confidence interval around the > tau estimate and check if estimates overlap? STATA reports the ASE > for tau with the tab y x, all command. > > HTH, > John Antonakis > > > On 04.12.2008 12:34, Marcus Fischer wrote: > > I would like to compare two Kenadalls' tau coefficients. Is it > possible to test whether one coefficient is significantly higher > than the other one ? > > Thanks. > > > > > * > * For searches and help try: > * http://www.stata.com/help.cgi?search > * http://www.stata.com/support/statalist/faq > * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
In reply to this post by Marcus Fischer
Yes, it is possible. However, how to do it depends on whether the
Kendall's tau coefficients are from the same sample, or from different samples. In Stata, confidence intervals for Kendall's tau-a statistics and their differences can be calculated using the -somersd- package, downloadable from SSC using the -ssc- command. This package is described in Newson (2002), Newson (2006a) and Newson (2006b), and calculates confidence intervals for a wide range of rank statistics, including some with sample weighting, clustering and/or stratification. Often, given a single outcome variable Y and 2 positive predictors X and W, we want to know which of the 2 positive predictors predicts Y better. This is because, as proved in Newson (2006a) and Newson (2002), if a positive association of Y with X is caused by a positive association of both variables with W, then we must have the inequalities tau(Y,X) <= tau (W,X) tau(X,Y) >= tau(W,Y) where tau(.,.) is Kendall's tau-a. Typically, in Stata, we type somersd Y X W, taua tdist lincom X-W The -somersd- command, with the -taua- option, calculates confidence intervals for the Kendall's tau-a parameters tau(Y,X) and tau(Y,W). The -lincom- command calculates a confidence interval for the difference tau(Y,X)-tau(Y,W). Sometimes, by contrast, we may want to estimate the difference between 2 tau(Y,X) statistics in different subpopulations, from which we have taken 2 different subsamples. In this case, we can use the -somersd- package with the -parmest- package, also downloadable from SSC. The -parmest- package creates and processes output datasets (or resultssets) with 1 observation per estimated parameter and data on estimates, standard errors, confidence limits and P-values, and is described in Newson (2003), Newson (2004), and Newson(2008). If the variable -subpop- indicates which of 2 subpopulations an observation belongs to, then we may type in Stata parmby "somersd Y X, taua tdist", by(subpop) norestore list metaparm [iwei=(subpop==2)-(subpop==1)] if parm=="X", list(,) The -parmby- command creates, in the memory, an output dataset (or resultsset), with 1 observation per fitted tau-a parameter and data on their estimates, standard errors, confidence limits, and P-values. (This resultsset overwrites the original dataset.) The tau-a parameters, are tau(Y,Y) in subpopulation 1, tau(Y,X) in subpopulation 1, tau(Y,Y) in subpopulation 2, and tau(Y,X) in subpopulation 2. We want to compare the 2 tau(Y,X) parameters. (The tau(Y,Y) parameters are simply probabilities that 2 randomly-chosen Y-values are unequal.) The list command lists all the estimated parameters. The metaparm command calculates and lists a confidence interval for the difference between the tau(Y,X) parameter in subpopulation 2 and the tau(Y,X) parameter in subpopulation 1. I hope this helps. Let me know if you have any queries. Best wishes Roger References Newson R. B. parmest and extensions. Presented at the 14th UK Stata User Meeting, 8-9 September, 2008. Download presentation from http://www.imperial.ac.uk/nhli/r.newson/usergp.htm Newson R. 2006b. Confidence intervals for rank statistics: Percentile slopes, differences, and ratios. The Stata Journal 6(4): 497-520. Buy the article from http://www.stata-journal.com/archives.html or download a pre-publication draft from http://www.imperial.ac.uk/nhli/r.newson/papers.htm Newson R. 2006a. Confidence intervals for rank statistics: Somers' D and extensions. The Stata Journal 6(3): 309-334. Buy the article from http://www.stata-journal.com/archives.html or download a pre-publication draft from http://www.imperial.ac.uk/nhli/r.newson/papers.htm Newson R. From datasets to resultssets in Stata. Presented at the 10th UK Stata User Meeting, 28-29 June, 2004. Download presentation and handouts from http://www.imperial.ac.uk/nhli/r.newson/usergp.htm Newson R. 2003. Confidence intervals and p-values for delivery to the end user. The Stata Journal 3(3): 245-269. Download the article from http://www.stata-journal.com/archives.html or download a pre-publication draft from http://www.imperial.ac.uk/nhli/r.newson/papers.htm Newson R. 2002. Parameters behind "nonparametric" statistics: Kendall's tau, Somers' D and median differences. The Stata Journal 2(1): 45-64. Download the article from http://www.stata-journal.com/archives.html Roger B Newson BSc MSc DPhil Lecturer in Medical Statistics Respiratory Epidemiology and Public Health Group National Heart and Lung Institute Imperial College London Royal Brompton Campus Room 33, Emmanuel Kaye Building 1B Manresa Road London SW3 6LR UNITED KINGDOM Tel: +44 (0)20 7352 8121 ext 3381 Fax: +44 (0)20 7351 8322 Email: [hidden email] Web page: http://www.imperial.ac.uk/nhli/r.newson/ Departmental Web page: http://www1.imperial.ac.uk/medicine/about/divisions/nhli/respiration/pop genetics/reph/ Opinions expressed are those of the author, not of the institution. -----Original Message----- From: [hidden email] [mailto:[hidden email]] On Behalf Of Marcus Fischer Sent: 04 December 2008 11:35 To: [hidden email] Subject: st: significance test between two kendalls' tau coefficents ? I would like to compare two Kenadalls' tau coefficients. Is it possible to test whether one coefficient is significantly higher than the other one ? Thanks. * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
In reply to this post by JOHN ANTONAKIS
On 4 Dec 2008, at 13:39, JOHN ANTONAKIS wrote:
> Why don't you simply construct a 95% confidence interval around the > tau > estimate and check if estimates overlap? STATA reports the ASE for tau > with the tab y x, all command. Warning: overlapping confidence intervals do not necessarily mean that the two are not significantly different. You need the confidence interval for the difference. Ronan Conroy ================================= [hidden email] Royal College of Surgeons in Ireland Epidemiology Department, Beaux Lane House, Dublin 2, Ireland +353 (0)1 402 2431 +353 (0)87 799 97 95 +353 (0)1 402 2764 (Fax - remember them?) http://rcsi.academia.edu/RonanConroy P Before printing, think about the environment * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
In reply to this post by Marcus Fischer
A >Date: Thu, 4 Dec 2008 16:13:01 +0000 >From: Ronan Conroy <[hidden email]> >Subject: Re: st: significance test between two kendalls' tau coefficents ? > >On 4 Dec 2008, at 13:39, JOHN ANTONAKIS wrote: > >> Why don't you simply construct a 95% confidence interval around the >> tau >> estimate and check if estimates overlap? STATA reports the ASE for tau >> with the tab y x, all command. > >Warning: overlapping confidence intervals do not necessarily mean that >the two are not significantly different. > >You need the confidence interval for the difference. Moreover, if you truly want a test as opposed to a CI, it is worth keeping in mind that the SE of the sampling distribution, and in fact all aspects of the sampling distribution, is in general not the same under the null (test situation) as it is under the alternative (CI situation.)I have seen cases with measures like Tau in which the two sampling distributions differed radically. An instructive exercise is to compare the distribution one gets from -bootstrap- (distribution under the alternative) vs. -permute- (under the null) Regards, Mike Lacy Fort Collins CO * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
Mike has raised an important issue here, because a "significant" difference between 2 Kendall's taus seems to imply that at least one of them is non-zero. A possible response is to use the hyperbolic arctangent transformation (also known as Fisher's z transformation), recommended by Edwardes (1995), as a Normalizing and variance-stabilizing transformation. This transformation is available in Stata by using the -somersd- package with the -transf(z)- option. If this is used, then -lincom- and -metaparm- will give confidence intervals for the difference between the z-transformed Kendall's taus, instead of confidence intervals for the difference between the original Kendall's taus. However, both differences have the same sign. It is also possible to use an arcsine transformation, as suggested by Daniels and Kendall (1947).
I hope this helps. Best wishes Roger References Daniels, H. E., and M. G. Kendall. 1947. The significance of rank correlation where parental correlation exists. Biometrika 34: 197–208. Edwardes, M. D. d. B. 1995. A confidence interval for Pr(X < Y) − Pr(X > Y) estimated from simple cluster samples. Biometrics 51: 571–578. Roger B Newson BSc MSc DPhil Lecturer in Medical Statistics Respiratory Epidemiology and Public Health Group National Heart and Lung Institute Imperial College London Royal Brompton Campus Room 33, Emmanuel Kaye Building 1B Manresa Road London SW3 6LR UNITED KINGDOM Tel: +44 (0)20 7352 8121 ext 3381 Fax: +44 (0)20 7351 8322 Email: [hidden email] Web page: http://www.imperial.ac.uk/nhli/r.newson/ Departmental Web page: http://www1.imperial.ac.uk/medicine/about/divisions/nhli/respiration/popgenetics/reph/ Opinions expressed are those of the author, not of the institution. -----Original Message----- From: [hidden email] [mailto:[hidden email]] On Behalf Of Mike Lacy Sent: 05 December 2008 18:30 To: [hidden email] Subject: Re: st: significance test between two kendalls' tau coefficents ? A >Date: Thu, 4 Dec 2008 16:13:01 +0000 >From: Ronan Conroy <[hidden email]> >Subject: Re: st: significance test between two kendalls' tau coefficents ? > >On 4 Dec 2008, at 13:39, JOHN ANTONAKIS wrote: > >> Why don't you simply construct a 95% confidence interval around the >> tau >> estimate and check if estimates overlap? STATA reports the ASE for tau >> with the tab y x, all command. > >Warning: overlapping confidence intervals do not necessarily mean that >the two are not significantly different. > >You need the confidence interval for the difference. Moreover, if you truly want a test as opposed to a CI, it is worth keeping in mind that the SE of the sampling distribution, and in fact all aspects of the sampling distribution, is in general not the same under the null (test situation) as it is under the alternative (CI situation.)I have seen cases with measures like Tau in which the two sampling distributions differed radically. An instructive exercise is to compare the distribution one gets from -bootstrap- (distribution under the alternative) vs. -permute- (under the null) Regards, Mike Lacy Fort Collins CO * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
In reply to this post by Roger Newson
Roger Newson wrote:
----------------------------------------------------------------------- Sometimes, by contrast, we may want to estimate the difference between 2 tau(Y,X) statistics in different subpopulations, from which we have taken 2 different subsamples. In this case, we can use the -somersd- package with the -parmest- package, also downloadable from SSC. The -parmest- package creates and processes output datasets (or resultssets) with 1 observation per estimated parameter and data on estimates, standard errors, confidence limits and P-values, and is described in Newson (2003), Newson (2004), and Newson(2008). If the variable -subpop- indicates which of 2 subpopulations an observation belongs to, then we may type in Stata ----------------------------------------------------------------------- But what about the case of clustering? In this case, there may be two distinct subsets of data on which tau(X, Y) is estimated, but there is a common pool of subjects. So, for example, Subject 1 contributed to tau(X ,Y) on both subsets. The within-data comparison of tau's in -somersd- won't work, because it only applies to common observations in the data set. The between-sets comparison is invalid as it ignores the clustering due to subjects. Al Feiveson * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
Al states, correctly, that -somersd- does not yet calculate confidence
limits for the difference between 2 within-stratum tau(X,Y) parameters specific to 2 strata that may both appear in the same cluster. To make this possible, I would probably have to upgrade -somersd- to create an output dataset (or resultsset), with 1 observation per cluster and data on the delta-jackknife influence functions or pseudovalues of that cluster for the various tau-a or Somers' D statistics. If we created 2 such resultssets, one for each stratum, then we could merge the 2 resultssets using the cluster() variable as the key, and then use the -mean- command to estimate a covariance matrix for the tau-a or Somers' D statistics, and then use -lincom- to estimate the difference. Such a method would be a rank-methods analog of -suest-. Best wishes Roger Roger B Newson BSc MSc DPhil Lecturer in Medical Statistics Respiratory Epidemiology and Public Health Group National Heart and Lung Institute Imperial College London Royal Brompton Campus Room 33, Emmanuel Kaye Building 1B Manresa Road London SW3 6LR UNITED KINGDOM Tel: +44 (0)20 7352 8121 ext 3381 Fax: +44 (0)20 7351 8322 Email: [hidden email] Web page: http://www.imperial.ac.uk/nhli/r.newson/ Departmental Web page: http://www1.imperial.ac.uk/medicine/about/divisions/nhli/respiration/pop genetics/reph/ Opinions expressed are those of the author, not of the institution. -----Original Message----- From: [hidden email] [mailto:[hidden email]] On Behalf Of Feiveson, Alan H. (JSC-SK311) Sent: 08 December 2008 18:45 To: [hidden email] Subject: st: RE: RE: significance test between two kendalls' tau coefficents ? Roger Newson wrote: ----------------------------------------------------------------------- Sometimes, by contrast, we may want to estimate the difference between 2 tau(Y,X) statistics in different subpopulations, from which we have taken 2 different subsamples. In this case, we can use the -somersd- package with the -parmest- package, also downloadable from SSC. The -parmest- package creates and processes output datasets (or resultssets) with 1 observation per estimated parameter and data on estimates, standard errors, confidence limits and P-values, and is described in Newson (2003), Newson (2004), and Newson(2008). If the variable -subpop- indicates which of 2 subpopulations an observation belongs to, then we may type in Stata ----------------------------------------------------------------------- But what about the case of clustering? In this case, there may be two distinct subsets of data on which tau(X, Y) is estimated, but there is a common pool of subjects. So, for example, Subject 1 contributed to tau(X ,Y) on both subsets. The within-data comparison of tau's in -somersd- won't work, because it only applies to common observations in the data set. The between-sets comparison is invalid as it ignores the clustering due to subjects. Al Feiveson * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/ |
Free forum by Nabble | Edit this page |