.-
help for ^eba^                                          (Gimpavido@@worldbank.org)
.-

Extreme Bound Analysis 
----------------------------

	^eba^ var1 var2 varlist [^if^ exp] [^in^ range] 
		, [^x(^varlist^) ty^pe^(^#^) le^vel^(^#^) vif(^#^) ci(^#^) d^etails]


Description
-----------

^eba^ Performs an Extreme Bound Analysis on the regressor "var2",
	for given a dependent variable "var1", and a set of regressors 
	Z and X. Where Z(nxk) and X(nxm) with k<=4.
	The program runs n!/(k!(n-k)!) OLS regressions by
	taking combinations of k<=4 Z variables among the p listed. It then
	displays minimum and maximum parameter estimates for "var2" together 
	with their t-statistics, p-value and Z variables used. This version
	controls for multicollinearity through -vif.ado- already coming
	with stata.

References: 	Levine R. and D. Renelt (1992)
		A sensitivity analysis of cross-country growth regressions.
		American Economic Review, 82(4), 942-63


Options
-------

^x(^varlist^)^ contains other possible regressors, common to all regressions.
	If it is not specified, it is assumed to be empty.

^type(^#^)^ tells the program how many Z variables should be taken for each
	combination (maximum 4). If it is not specified, it is assumed to be 1.

^level(^#^)^ tells the program within what confidence level the minimum and
	maximum for "ebavar" should be reported. If it is not specified, 
	it is assumed to be 0.95.

^vif(^#^)^ controls for the collinearity problem. It sets the maximum
	admissible VIF below which extremes have to be found. If it is not
	specified, the maximum is by default very high: i.e. it allows for 
	a high collinearity in the auxiliary regressions.
	N.B. 	VIF is calculated for each of the n!/(k!(n-k)!) regressions
		and it slows down the program considerably. Hope to solve this
		sometime.

^ci(^#^)^ modyfies the confidence for the intervals of both the minimum and 
	maximum for "ebavar". If it is not specified, it is assumed to be .95.

^detail^ displays parameter estimate, t-statistics, p-value, maximum VIF
	and relative Z 	variables for all regressions performed.


Examples
--------

	. ^eba y x1 x2 x3^
	. ^eba y x1-x3, x(x4-x6)^
	. ^eba y x1 x2 x3 x4 x5, type(4)^
	. ^eba y x1 x2 x3 x4 x5, x(x6) type(4) detail le(.85) ci(.90)^
	. ^eba y x1 x2 x3 x4 x5, x(x6) type(4) detail le(.85) vif(10) ci(.90)^


Also see
--------

 Manual:  ^[R] fit^
On-line:  help for @fit@


Author
------

	Gregorio Impavido
	DECVP
	The World Bank
	Washington, DC
	Gimpavido@@worldbank.org