{smcl}
{* 27Oct2005}{...}
{hline}
help for {hi:plreg}{right:(SJ6-3: st0109)}
{hline}

{title:Linear partial regression }

{p 8 17 2}{cmd:plreg} {it:depvar} {it:indepvars} {ifin}{cmd:,}
{opth nlf(varname)}
[{opth gen:erate(newvar)} {opt or:der(#)}
	{cmd:wh}
	{opt l:evel(#)}
	{opt coll:inear}
	{it:lowess_options}]


{title:Description}

{p 4 4 2}
{cmd:plreg} estimates the following semiparametric regression model by the
method of differencing:

        Y = f(z) + x*b + e          (1)

{p 4 4 2}
where f(z) is a smooth function with bounded first derivatives,
the function f is known to lie in a particular parametric family,
x are control variables that enter (1) linearily, e is the zero-mean
innovation error, and b is a vector of parameters. {p_end}

{p 4 4 2}
Standard errors of b are adjusted according to Yatchew (1998) method.{p_end}


{title:Options}

{p 4 8 2}
{opth nlf(varname)} is required and specifies the argument of an unknown
function.

{p 4 8 2}
{opth generate(newvar)} creates a new variable {it:newvar} containing the
smoothed value of {it:f}. These values is estimated by the locally weighted
regression using {cmd:lowess}. A corresponding graph of the estimated function {it:f} could also be output; see {helpb lowess}.

{p 4 8 2}
{cmd:order(}{it:#}{cmd:)} specifies the differencing order. Tenth-order
differencing is the maximum allowed.  If {cmd:order(}{it:#}{cmd:)} is not
specified, the model is fitted by first-order differencing.

{p 4 8 2}
{cmd:wh} specifies a form of the vector of differencing weights. By default,
Yatchew (1998) weights are used. If {cmd:wh} is specified, Hall, Kay, and
Titterington (1990) weights are used for differencing.

{p 4 8 2}
{cmd:level(}{it:#}{cmd:)} sets the confidence level; the default is
{cmd:level(95)}.

{p 4 8 2}
{cmd:collinear} specifies that collinear variables be kept.

{p 4 8 2}
{it:{help lowess:lowess_options}} control the way {cmd:lowess} generates the
smoothed values for the argument.


{title:Examples}

{p 4 8 2}
{cmd:. plreg y1 x1 x2 x3, nlf(z) }

{p 4 8 2}
{cmd:. plreg y1 x1 x2 x3, nlf(z) wh gen(f_hat) order(7)}

{p 4 8 2}
{cmd:. plreg y1 x1 x2 x3, nlf(z) wh gen(f_hat) order(7) nodraw}


{title:References}

{p 4 8 2}
Hall, P., J. Kay, and D. Titterington. 1990. Asymptotically optimal
difference-based estimation of variance in non-parametric regression.
{it:Biometrica} 77: 521-528.

{p 4 8 2}
Robinson, P. 1988. Root-n-consistent semi-parametric regression.
{it:Econometrica} 56: 931-54.

{p 4 8 2}
Yatchew, A. 1998. Nonparametric regression techniques in economics.
{it:Journal of Economic Literature} 36: 669-721.


{title:Author}

{pstd}
 M. Lokshin, DECRG, The World Bank.{break}
 If you observe any problems {browse "mailto:mlokshin@worldbank.org"}.


{title:Also see}

{p 4 13 2}
Online:  {helpb regress}, {helpb lowess}
{p_end}