Name

auto_stage1 — search of a statistically admissible regression

CALLING SEQUENCE

[listeq]=auto_stage1(y,x,listeq,indx,indcte,firstelim,r00,rmod,alpha,eta,z)

PARAMETERS

Input

• y = vector of the endogenous variable

• x = matrix of the exogenous variables

• listeq = the list containing all preceding paths and the corresponding estimation results

• index = the indexes of the x variables remaining at the entry of the function

• indcte = the booelan indicating the presence or the absence of the constant in the regression

• firstelim = the index of the first variable which is withdrawn to begin the path

• r00 = the results tlist for the regression containing all exogenous variables whose index is in indx

• rmod = a tlist defined by def_results

• alpha = broad significance level (typically lower than f0_sig)

• eta = (px1) significance level for specification tests

• z = matrix of the compulsory variables (the ones that must be in the regression whatever significance they have)

Output

• listeq = the entry list plus the path followed here and, if it leads to a new model, the corresponding estimation results

DESCRIPTION

Starting from the variable whose index in indx is firstelim, find a path where: the less significant variable which does not make the specification tests calculated in test_func fails is successively withdrawn until: 1) either the withdrawn variables as a group has been already encountered on another path 2) or all remaining variables are non significant or the regression does not pass the specification tests calculated in test_func.

EXAMPLE

This function is a procedure of automatic. Very specific indeed.

               

AUTHOR

Eric Dubois 2002