help parttau


parttau -- Calculate Kendall's partial tau and confidence interval.

Syntax parttau varlist [if] [in] [, options]

When varlist contains three or more variables, the first two are the primary variables, and all others confounders. When varlist contains two variables, the pairwise tau is calculated.

options Description ------------------------------------------------------------------------- Main cluster(varname) treat data set as samples of clusters wcluster calculate the within-cluster partial and pairwise taus estimate(varname) save jackknife estimates in given variable name transform(string) report the tau, SE, and confidence interval in a transformed space; transformations are same as for somersd; see below level(#) specifies percent coverage for confidence interval run2 forces calculation of pairwise tau when no confounder is given; see below show displays all pairwise taus between all given variables -------------------------------------------------------------------------

Options +------+ ----+ Main +-------------------------------------------------------------

cluster(varname) instructs the program that the data set should be treated as samples of clusters from a population of clusters, and that the given variable specifies the clusters. If cluster is specified and the option wcluster is not specified, then the partial tau is calculated between clusters.

estimate(varname) specifies that the jackknife estimates should be saved in variable varname. If the variable exists prior to calling this option, the variable is overwritten and a warning message given. Otherwise a variable with the given name is created.

transform(name) changes the ouput estimate, SE, and confidence interval to a transformed space. If specified, the confidence interval will be symmetric in the transformed space. Allowed transformations are same as for somersd: z, asin, rho, zrho, c, though fewer aliases are allowed. If calculating a partial tau (i.e., using one or more confounder variables), and option show is specified, reported pairwise taus will not be transformed.

level(#) specifies the percent coverage of the confidence interval.

run2 forces calculation of pairwise tau. parttau is primarily intended for calculating Kendall's partial tau between two variables with one or more confounders. Thus, typically, three or more variables will be specified. If two variables are specified, parttau will assume that the user wants to calculate the pairwise tau and will attempt to use somersd. If that program is not found, parttau will calculate the pairwise tau itself. Using run2 forces parttau to calculate the pairwise tau; if somersd is installed, both will be used.

show displays all pairwise taus between all given variables. The pairwise taus are also saved in the matrix e(taus).


parttau calculates Kendall's partial tau and confidence interval. parttau can be used to calculate pairwise taus if somersd is not installed, or if the option run2 is specified. For repeated calculation of the pairwise tau, the user will probably find somersd preferable; it is faster and has more options.


James Fiedler, Universities Space Research Association Email:

Alan H. Feiveson, National Aeronautics and Space Administration