<P>请教<STRONG><FONT face=Verdana>arlionn,用xtscc这个命令如何纠正固定效应模型的组间、组内异方差和一阶自相关?</FONT></STRONG></P>
<P><STRONG><FONT face=Verdana>我看了那个文献和HELP命令,但还是理解不透彻!</FONT></STRONG></P>
<P><STRONG><FONT face=Verdana>help上写的例子是,</FONT></STRONG></P>
<P> . est clear</P>
<P> . xtreg invest mvalue kstock, fe robust</P>
<P> . est store fe_robust</P>
<P> . xtscc invest mvalue kstock, fe lag(4)</P>
<P> . est store fe_dris_kraay</P>
<P> . est table *, b se t</P>
<P>这里的LAG(4)是不是表示纠正四阶自相关?那么一阶自相关是不是用lag(1)就可以?</P>
<P>如果直接输入xtscc invest mvalue kstock, fe lag(4)去估计不可以么?</P>
<P>如果用命令xtscc invest mvalue kstock, fe lag(1)是否就可以纠正异方差和自相关了?</P>
<P>还有XTTEST3检验出来的是组内异方差吧?</P>
Therefore, the xtscc program uses a simple rule of thumb for selecting m(T ) when no lag(#) option is specified. The heuristic applied is taken from the first step of Newey and West’s (1994) plug-in procedure and sets
m(T ) = floor[4(T/100)2/9] .
Note, however, that choosing the lag length like this is not necessarily optimal because this choice is essentially independent from the underlying data. In fact, this simple rule of selecting the lag length tends to choose an m(T ) which might often be too small.
这面是stata journal 2007, vol7-3, pp.281-312 原文中的一段话,我的理解是如果你不设定lag(#),它就会自动根据公式计算,到底lag多少,应该不是很重要吧。