<< pdccemg_jkn1 Panel equation regressions pfixed1 >>

Grocer >> Panel equation regressions > pfixed

pfixed

Panel equation regressions

CALLING SEQUENCE

res=pfixed(namey,arg1,...,argn)

PARAMETERS

Input

* namey = a real (nx1) vector or a string equal to the name of a time series or a (nx1) real vector between quotes (this last case is the only one authorized if you are using a 'panel data' tlist, see below)

* arg1 =

  - either a 'panel data' tlist (generally imported from a .csv database by function impexc2bd)

  - or an endogenous variable taking the form of a time series, a real (nx1) vector or a string equal to the name of a time series or a (nx1) real vector between quotes

* arg2,...,argn =

  - if first input of arg1,...,argn was a 'panel data' tlist then: other input are optional and can be:

    . either 'x=name1;...;namep' where name1,...,namep are a subset of the names of the variables that are in the database

    . the string 'nameid=name1,..., namen' where name1,...,namen are names of individuals present in the database

    . or the string 'noprint' if the user does not want to print the estimation results

  - if first input of arg1,...,argn was an endegnous variable then either:

    . a time series

    . a real (nxk) matrix

    . a (k x 1) string vector of names of time series, vectors or matrices

    . the string 'id=v' where v is the vector of individuals attached to the y and x data (this argument must be present somewhere in the list of variables arguments)

  - 'hac=ccm' for "clustered" covariance matrix of Arellano (1987) recommended when T is fixed and N large but "works" also when T is large and N fixed see Hansen C. B. (2007) (reference below) or 'hac=nw' for a Newey-west type estimator recommended when T is large and N fixed (see references) or 'hac=mbb' for a moving block bootstrap etimation of the symetric percentile-t confidence interval of the betas

  - 'nboot=xxx' number of bootstrap replications when 'hac=mbb' (default=999)

  - 'alpha=xxx' with 0<xxx<1 the confidence level (default=0.95) for the symetric percentile-t confidence interval

  - 'win=xxx' the length of the Barlett window kernel estimator (default = automatic selection by Andrews (1991) method using an AR(1) model)

  - the string 'noprint' if the user doesn't want the to print the results of the regression

 

Output

* res = a tlist with

  -res('meth') = 'panel with random effects'

  -res('y') = y data vector

  -res('x') = x data matrix

  -res('nobs') = nobs

  -res('nvar') = nvars

  -res('beta') = bhat

  -res('yhat') = yhat

  -res('resid') = residuals

  -res('vcovar') = estimated variance-covariance matrix of beta

  -res('sige') = estimated variance of the residuals

  -res('sige') = estimated variance of the residuals

  -res('ser') = standard error of the regression

  -res('tstat') = t-stats

  -res('pvalue') = pvalue of the betas

  -res('condindex') = multicolinearity cond index

  -res('prescte') = boolean indicating the presence or absence of a constant in the regression

  -res('rsqr') = rsquared

  -res('rbar') = rbar-squared

  -res('f') = F-stat for the nullity of coefficients other than the constant

  -res('pvaluef') = its significance level

DESCRIPTION

Performs HAC or standard fixed effects estimation for panel data (for balanced or unbalanced data).

EXAMPLE

load(GROCERDIR+'\macros\grocer\db\judgepanel.dat') ;
r2 = pfixed('y',judgepanel);
// Example is taken from function panel_d. Provides fixed panel estimation on Judge et alii example.

AUTHOR

Eric Dubois 2005 and Emmanuel Michaux 2012

Report an issue
<< pdccemg_jkn1 Panel equation regressions pfixed1 >>