Code: Select all
equation eq0.ls(cx=f, per=f) i c f
equation eq1.ls i c f @expand(@crossid, @dropfirst) @expand(@year, @dropfirst)Thanks Glenn for your reply. However, I did not fully grasp your answer. Basically I am teaching a course and give an exercise to replicate fixed effect model results by using eviews panel options and dummy variables. They did for cross section and period separately. However, they and as well as I did not able to match results if using both cross-section and period fixed effects. Since, I am using n-1 cross section dummies and k-1 period dummies then constant have impact of a period and cross section. And I don't know how to separate these effect. I need your help in this regard and will highly appreciate.In an annual workfile, something likewill match the results for the regressors. You'll have to do additional work to get the dummies to match since we estimate a different normalization of the constants.Code: Select all
equation eq0.ls(cx=f, per=f) i c f equation eq1.ls i c f @expand(@crossid, @dropfirst) @expand(@year, @dropfirst)
Code: Select all
' The following shows one method for normalizing direct dummy variables estimates to match the EViews two-way effects normalization
' to perform the computations, you should open the grunfeld_baltagi_panel.wf1 workfile provided with your copy of EViews. Note that the validity
' of the computations below assume that there is a balanced panel.
' estimate with dummy variables leaving out one year category
equation eqdirect.ls f i @expand(fn) @expand(yr, @dropfirst)
' estimate with built-in fixed effects
equation eqeviews.ls(cx=f,per=f) f i
' extract the full set of coefficients from the dummy variables regression
vector directcoefs = eqdirect.@coefs
' extract just the fixed effects estimates from the dummies
vector directcx = @subextract(directcoefs,2,1,11,1) ' cross-section effects
vector directper = @subextract(directcoefs,12,1,eqdirect.@ncoefs,1) ' period effects
' compute the mean of the cross-section effects and center those effects
scalar overall_mean = @mean(directcx)
vector cxeff = directcx - overall_mean ' center the effects
' compute the mean of the period effects
scalar per_mean = @sum(eq1yr)/20
' create full vector of period effects putting a zero for the omitted category then center the effects
vector(20) pereff
pereff(1) = 0
matplace(pereff, directper, 2,1)
pereff = pereff - per_mean ' center the effects
' compute overall constant from the centering means
scalar adjusted_mean = overall_mean + per_mean
' grab the intercept and the beta
vector(2) coefs
coefs(1) = adjusted_mean
coefs(2) = directcoefs(1)
' you may compare with the results from the equation object EQEVIEWS