{smcl}
{* Mars 2012}{...}
{hline}
help for {hi:pcmtest}{right:JFH}
{hline}

{title:Global and item specific tests of the fit for Partial Cradit Model}

{p 8 14 2}{cmd:pcmtest}, [{cmdab:g:roup}({it:numlist})
{cmdab:n:fit}(#)
{cmdab:p:ower}(#)
{cmdab:graph:ics}(#)
{cmdab:a:lpha}(#)
{cmdab:ap:proximation}
{cmdab:n:ew}]

{title:Description}

{p 8 14 2}{cmd:pcmtest} allows to test the fit between an observed data and the
 partial credit model.{p_end}
{p 14 14 2}The global test of the fit (test R1m) can only be computed if the data
 contains at least two items. Item specific test of the fit (Si tests) can only be
  computed if the data contains at least three items. {p_end}
{p 14 14 2}{cmd:pcmtest} is only performed if item difficulties have been
  estimated using {cmd:pcmodel}. {cmd:pcmtest} can be performed either if covariates
   have been included in {cmd:pcmodel} or not.

{title:Options}

{p 4 8 2}{cmd:group} specifies groups of scores, by defining the superior limits of
 each group

{p 4 8 2}{cmd:nfit} specifies the size of a virtual sample with items response
 distribution strictly identical to that observed ({cmd:nfit} can deal with
  over-power problems when tests of the fit are performed on large samples).

{p 4 8 2}{cmd:power} specifies the desired power for a test R1m, and estimates the
 corresponding size of a virtual sample with items response distribution strictly
  identical to that observed.

{p 4 8 2}{cmd:alpha} specifies the type I error used to performe the tests of the fit
 (default: 0.05).

{p 4 8 2}{cmd:graphics} displays several graphs (Graphical fit between the observed and
 expected latent traits according to the individual scores, graphic of MAP, graphic of
  groups contribution to the R1m statistic, and graphice of the observed and expected
   score distribution)

{p 4 8 2}{cmd:approximation} Computation of the pattern response probabilities using 
simulation instead of Gauss Hermitte quadratures. Computation should be quicker for 
high number of items or responses modalities.

{p 4 8 2}{cmd:new} Do not use pattern response probabilities computed during a previous 
test of the fit, compute it an other time! (allows changing the computation methodology)

{title:Author}

{p 4 8 2}Jean-François HAMEL{p_end}

{title:Also see}

{p 4 13 2}Online: help for {help pcmodel}, {help gllamm}, {help simirt}
, {help raschtest}.{p_end}