st: RE: AW: nlcom

["Nick Cox" <n.j.cox@durham.ac.uk>]

Without attempting to answer this directly I did notice that in your last command the reference to 

exp(xb1) 

was probably a typo for 

exp(`xb1') 

Nick 
n.j.cox@durham.ac.uk 

Martin Weiss

Marginal effects are best estimated using -margins-. What is your reason for
not using it? 

What is the error description (the code number is not that informative)?

If you want -quietly- and the -r(mean)- returned value, you can also use the
-meanonly- option of -summ-.

Olivia Bacha

I am running the following code after logit estimation using nlcom,
this is to estimate marginal effects.


quietly sum gastos
local gastosM= r(mean)

quietly sum var60
local var60M= r(mean)

quietly sum gastopc
local gastopcM= r(mean)

quietly sum gastopcreeleição
local gastopcreeleiçãoM=r(mean)

quietly sum Idade
local IdadeM=r(mean)

quietly sum qualidade
local qualidadeM=r(mean)

quietly sum qtpart
local qtpartM=r(mean)

quietly sum educ
local educM=r(mean)

local xb1 _b[prop_gasto]*(`gastosM'/`var60M') + _b[gastopc]*`gastopcM'
+ _b[reeleição] + _b[gastopcreeleição]*`gastopcreeleiçãoM' +
_b[Idade]*`IdadeM' + _b[qualidade]*`qualidadeM' + _b[qtpart]*`qtpartM'
+ _b[educ]*`educM'


local xb0 _b[prop_gasto]*(`gastosM'/`var60M') + _b[gastopc]*`gastopcM'
+ _b[Idade]*`IdadeM' + _b[qualidade]*`qualidadeM' +
_b[qtpart]*`qtpartM' + _b[educ]*`educM'



nlcom (a1:(1/(1+ exp(`xb1')))^2) (a2:(`var60M' - `gastosM')/`var60M'^2),
post

nlcom _b[a1] * _b[prop_gasto] * _b[a2]


nlcom (-1) * ((1/(1+ exp(`xb0'))) ^2) * _b[prop_gasto] * (`var60M' -
`gastosM')/`var60M'^2


 nlcom ((1/(1+ exp(xb1))) ^2) * _b[prop_gasto] * (1/`var60M')^2 * `gastosM'


Stata is returning error 198 when i start using nlcom. What's my mistake?


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/