help for ^partgam^   Partial Gamma        26. apr. 1999             

^partgam^ v1 v2 [^if^ exp] [^in^ range] [weight] [, ^sub^table ^t^able ^subp^ct ^adj^ust(varlist) ^by^(varlist) ^not^able ^test l^evel() ^one dl^() ^r^ef() ^i^nfo ] ^Description^ ---------- partgam Calculation and tests for partial gamma coefficient (stata 6)

Partgam is relevant for overall and stratified (conditional) analysis of two variables which are of type ^ordinal^. If called without any ^adjust^ variables, the procedure will show the gamma with confidence intervals for the overall table.

The partial gamma coefficient is a weighted summary version of gamma across subtables of the Goodman & Kruskal's gamma coefficient. A test of homogeneity of stratified gamma's is applied.

See ref. for gamma in Stata Manual under @tabulate@ 1: Kruskal & Wallis 2: Agresti

By default observations with missing (.) will be excluded. Variable labels are only shown frist time a particular variable is used in subsequent runs.

^Options^ -------- ^tab^le Show table of overall counts ^tabpct^ Show table of overall counts with row percentage ^sub^table Show stratified table ^subp^ct Show stratified table with row percentages ^not^able Do not show stratified confidence intervals for gamma

^adj^ust(var) Stratify (condition) on variable. Only one variable allowed Can be exchanged with ^by(var)^ ^m^iss Include all cases with missing on stratification variables ^one^ Calculate one-sided tests for gamma (default is two sided) ^i^nfo Write out all information on ^varlist, missing etc^ ^dl^ Set display and log line length (default 100) ^t^est Apply a test of whether gamma's are equal to reference Default reference category is the first stratum shown in the stratified results. See option ^ref^. ^r^ef Reference stratum, when ^test^ is applied (default 1) ^l^evel(int) Level for confidence interval (default $S_level, see @level@)

----------------------------------------------------------------------------- Files are: partgam.ado, partgam.hlp, partgam1.do, partgam2.do Latest version accessible as download by: Stata version 6 : net from http://www.bola.suite.dk Internet browser: point at http://www.bola.suite.dk As an STB contribution if accepted by STB

Additional information: It can be useful to execute @pattern@ STB 33) to see which variables have missing values (^search pattern,STB^). If you wish to recode continous variables into ordinal type variables use the function @cut@, see STB52.

When stratifying the following results are returned: ^matrix form^: r(ase) r(high) r(low) r(gamma) ^scalars^:r(strata) r(partgam) r(hi_gam) r(lo_gam) r(p_gamma) r(chi_gam) r(p_chi)

Note ^only one variable for adjust^ allowed. But you can yourself generate a variable group variable if you want to adjust for more than one variable: egen group=group(var1 var2 var3) partgam v1 v2, adjust(group)

Examples ---------

. ^partgam agegrp incgrp, adjust(sex)^ . ^partgam agegrp incgrp, adjust(sex) table sw level(90) one^ . ^partgam agegrp incgrp, adjust(sex) subpct table^ . ^partgam health socgrp, adjust(agegrp) subpct table one info all^

Applying if in weights . ^partgam d1 d2 if r > 0^ . ^partgam d1 d2 in 1/400^ . ^partgam d1 d2 [freq=pop] ^

Author ------- Jens M.Lauritsen, County of Fyn, Odense Denmark jm.lauritsen@@dadlnet.dk Svend Kreiner svend.kreiner@@sociology.ku.dk

Also see --------- On-line: help for @tabulate@ @pattern@ @level@