You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

87 lines
3.4 KiB
Plaintext

.-
help for ^countfit^ - 03Nov2005
.-
Compare fit of alternive count models
-------------------------------------
^countfit^ varlist [if] [in]^,^ [^inf^late^(^varlist2^)^ ^noc^onstant ^p^rm ^n^breg
^zip^ ^zin^b ^g^enerate^(^prefix^)^ ^replace^ ^note(^string^)^
^graph^export^(^filename[^, replace^]^)^ ^nog^raph ^nod^ifferences ^noprt^able
^noe^stimates ^nof^it ^nodash^ ^max^count(#) ^noi^sily]
Description
-----------
^countfit^ compares the fit of the Poissin, negative binomial, zero-inflated
Poisson and zero-inflated negative binomial models, generateing a table of
estimates, a table of differences between observed and average estimated
probabilities for each count, a graph of these differences, and various tests
and measures used to compare the fit of count models.
Specifying the model
--------------------
Immediately afer the command name ^countfit^ you specify the dependent and
independent variables as you would with ^poisson^ or the other models. For
zero-inflated models, the ^inflate^ option is used in the same was as in the
^zip^ and ^zinb^ commands. ^noconstant^ can be used to exclude the constant
term.
Options to select the models to fit
-----------------------------------
By default, ^poisson^, ^nbreg^, ^zip^ and ^zinb^ are estimated. If you
only want some of these models, specify the modesl you want with:
^prm^ - include ^poisson^
^nbreg^ - include ^nbreg^
^zip^ - include ^poisson^
^zinb^ - include ^poisson^
Options to label results
------------------------
^generate()^ is up to five letters to name the variables that are created
and to label the models in the output. This name is placed in front
of the type of model (e.g., namePRM). This is optional but will help
keep track of results from multiple specifications of models.
^replace^ will replace variables created with the ^generate^ option if they
already exist.
^note()^ is a label added to the graph that is saved.
^graphexport()^ contains options to be sent to the ^graph export^ command to
export the graph that is create. For example, ^graph(mdl.emf, replace)^
would save the file mdl.emf and replace it if it exists. If this option
is not included, the graph is not saved.
Options controlling what is printed
-----------------------------------
^noisily^ shows the output from the estimation commands called by ^countfit^.
^nograph^ suppress graph of differences from observed counts.
^nodifferences^ suppress table of differences from observed counts.
^noprtable^ suppresses table of predictions for each model.
^noestimates^ suppress table of estimated coefficients.
^nofit^ suppress table of fit statistics and test of fit
^nodash^ suppress dashed lines between measures of fit
^maxcount()^ number of counts to evaluate.
^noisily^ includes output from Stata estimation commands; without this option
the results are only shown in the ^estimates table^ output.
Notes
-----
^countfit^ is based on the results from the Stata models described above,
the predictions computed by ^prcounts^, and the fit measures computed by
^fitstat^.
Examples
--------
^. use couart2^
^. countfit art fem mar kid5 phd ment, inf(ment fem) nbreg zinb nograph^
.-
Author: J. Scott Long and Jeremy Freese
www.indiana.edu/~jslsoc/spost.htm
spostsup@@indiana.edu