{smcl}
{* 08jan2001}{...}
{hline}
help for {hi:graphf}{right:(STB-60: gr46)}
{hline}

{title:Plot the "total" effect of a variable}

{p 8 16}
{cmd:graphf} {it:varname} [{it:varlist}]
[{cmd:,}
  {cmd:eq(}{it:str}{cmd:)}
  {cmd:cb(}{cmdab:n:one}|{cmdab:e:nvelope}|{cmdab:v:alue}|{cmd:bs} [{it:#}]{cmd:)}
  {cmdab:l:evel}{cmd:(}{it:#}{cmd:)}
]


{title:Description}

{p}{cmd:graphf} is a post-estimation command that produces a graph of the
fitted values and a confidence band of the effect of {it:varname}, modeled
via a linear predictor in (usually nonlinear) transformations of
{it:varname}. For instance, if the effect of working experience ({hi:exp}) is
modeled via a linear term {hi:exp} and a quadratic term {hi:exp2} =
{hi:exp^2}, then {it:varname} = {hi:exp} and {it:varlist} = {hi:exp exp2}.
Variables in {it:varlist} that do not occur in the model are skipped with a
warning.

{p}If {it:varlist} is not specified, {cmd:graphf} will determine itself the
{it:varlist} as the variables in the (selected equation of) the model that
are a function of {it:varname}, i.e., do not vary within levels of
{it:varname}.


{title:Options}

{p 0 4}
{cmd:eq(}{it:str}{cmd:)} specifies the name of the equation in which the
variables occur.  If {cmd:eq()} is not specified, {cmd:graphf} uses the first
equation of the model.

{p 0 4}
{cmd:cb(}{it:value}{cmd:)} specifies the type of confidence band to be
  displayed.  Valid values are

{p 6 16}
{cmd:none}{space 5}
  no band is displayed (the default)

{p 6 16}
{cmd:value}{space 4}
  Specifies a value-wise confidence band that consists of the spline-connected
  lower and upper limits of the value-wise confidence intervals.

{p 6 16}
{cmd:envelope}{space 2}
  Ripley's envelope confidence band is displayed.  This band is obtained as the
  value-wise minimum and maximum of the fitted values from 19 draws from the
  estimated multivariate distribution of the estimators for the coefficients
  in the model estimated last.

{p 6 16}
{cmd:bs} [{it:#}]{space 3}
  A parametric bootstrap confidence band, assuming multivariate normality of
  the parameter estimates. The value-wise confidence interval is estimated
  by the bias-corrected method. The number of Monte Carlo simulations {it:#}
  defaults to 1000.

{p 0 4}
{cmd:level(}{it:#}{cmd:)} specifies the confidence level, in percent, for the
value-wise confidence intervals.  The default is {cmd:level(95)} or as set by
{help set level}.


{title:Examples}

{p 8 12}{inp:. gen exp2 = exp^2}{p_end}
{p 8 12}{inp:. regress inc edu exp exp2}{p_end}
{p 8 12}{inp:. graphf exp  exp exp2}{p_end}

{p 8 30}{inp:. graphf exp}{space 9}
{hi:graphf} determines that {hi:exp} and {hi:exp2} are vars included
in {cmd:regress} that are functions of {hi:exp}
{p_end}

{p 8 12}{inp:. mkspline e_1 1 e_2 5 e_3 10 e_4 = exp}{p_end}
{p 8 12}{inp:. regress inc edu e_1 e_2 e_3 e_4}{p_end}
{p 8 12}{inp:. graphf exp e_*, ci(v)}{p_end}
{p 8 12}{inp:. graphf exp, ci(bs 400)}{p_end}


{title:Author}

{tab}Jeroen Weesie
{tab}Dept of Sociology/ICS
{tab}Utrecht University
{tab}J.Weesie@fss.uu.nl


{title:Also see}

{p 0 21}On-line:  help for {help predict}, {help wherext}{p_end}