Dear all,
I am fitting a fixed-effects model using xtregar in Stata 13.0 on Win 10, and export the fitted data to a .csv file to create a 3-D graph in an external application (Matlab). The code below worked nicely until it suddenly stopped working. I don't recall having made any changes to the code, but only to the model specification. Yet, I receive a "conformability error" r(503);
Questions:
1. Can someone spot/correct the error ?
2. Are their alternative and possibly easier approaches to creating high(er)-quality 3-D graphs (see example below)? The "surface" library seems rather limited
Thank you for your help!
3D Graph created in Matlab:
I am fitting a fixed-effects model using xtregar in Stata 13.0 on Win 10, and export the fitted data to a .csv file to create a 3-D graph in an external application (Matlab). The code below worked nicely until it suddenly stopped working. I don't recall having made any changes to the code, but only to the model specification. Yet, I receive a "conformability error" r(503);
Questions:
1. Can someone spot/correct the error ?
2. Are their alternative and possibly easier approaches to creating high(er)-quality 3-D graphs (see example below)? The "surface" library seems rather limited
Thank you for your help!
Code:
eststo: quietly xtregar $dv c.$iv1 c.l.fsu c.$iv1#c.l.fsu c.$iv1#c.$iv1 $ctrl_a , fe
matrix b =e(b)
margins, atmeans post
matrix avg = e(at)
matrix list avg
scalar minx = 0.15 // set the minimum value for variable1
scalar maxx = 1 // set the maximum value for variable1
scalar intervalx = 0.05 // set the periodic interval for variable1 marginal estimation
scalar miny = 0 // set the minimum value for variable2
scalar maxy = 0.16 // set the maximum value for variable2
scalar intervaly = 0.01 // set the periodic interval for variable2 marginal estimation
scalar rows = (((abs(minx) + abs(maxx))/intervalx)+1)*(((abs(miny)+ abs(maxy))/intervaly)+1) //set the number of rows for the matrix to be iterated
matrix fitted = J(rows ,3,.) //generates the matrix to be iterated
capture drop i
gen i = 1
forvalues x = `=minx'(`=intervalx')`=maxx' {
forvalues y = `=miny'(`=intervaly')`=maxy' {
forv z = 1(1)2{ // selects the marginal variables from the regression
matrix b = b[1,`z'..`z']
global name`z': colname(b)
}
scalar numbv= colsof(b)-1 // selects the fixed variables from the regression
matrix b = b[1,3..numbv]
global fixedvars: colname(b)
eststo: quietly xtregar $dv c.$iv1 c.l.fsu c.$iv1#c.l.fsu c.$iv1#c.$iv1 $ctrl_a , fe
eststo: quietly margins, at($name1==(`x') $name2==(`y')) atmeans continuous vsquish post
scalar ypredict = _b[_cons]
matrix fitted[i,1] = ypredict
matrix fitted[i,2] = `x'
matrix fitted[i,3] = `y'
replace i = i + 1
estimates drop _all
}
}
drop i
preserve
// transforms the matrix into a dataset
matsave fitted, p("$main_analysis_dir") saving dropall replace
use "fitted.dta"
export delimited using "3DGrapg.csv", replace
surface c2 c3 c1, xtitle("Competitor exploration (t)") ytitle("Technological Proximity to Competitors (t)") ztitle("Firm exploration (t)") scheme(s1manual) // generates 3d graph
restore
3D Graph created in Matlab:
