Scatter of a local

[Iris Etchevarren Acquarone]

Hi, is it not possible to create a scatter of a local variable? I'm evaluating the difference in the probabilities of Y btn men and women for a range of other variables. Through some loops I get a list of that, but when trying to plot it, I can't. Is not possible to make a scatter of a local?

This code is working (with it I get the mentioned list):

quietly {

logit pmvotegrp lrself male retnat income age edulevel

local maleA=0
local incomeA=3
local lrselfA=5
local edulevelA=3
local retnatA=3

local maleB=1
local incomeB=3
local lrselfB=5
local edulevelB=3
local retnatB=3

capture drop predA
capture drop predB

foreach nn of numlist 1/100 {

local ageA=`nn'
local ageB=`nn'


# delimit ;
local sysA=_b[_cons]+
_b[lrself]*`lrselfA'+
_b[male]*`maleA'+
_b[retnat]*`retnatA'+
_b[income]*`incomeA'+
_b[age]*`ageA'+
_b[edulevel]*`edulevelA';

local sysB=_b[_cons]+
_b[lrself]*`lrselfB'+
_b[male]*`maleB'+
_b[retnat]*`retnatB'+
_b[income]*`incomeB'+
_b[age]*`ageB'+
_b[edulevel]*`edulevelB';

local predA=exp(`sysA')/(1+exp(`sysA'));
local predB=exp(`sysB')/(1+exp(`sysB'));

# delimit cr

local diffBA=`predB'-`predA'

noisily: display in r "Case `nn' (Age `ageB'): `diffBA' "

}
}

When changing the noisily part for what follows, I get nothing:

noisily: twoway scatter `diffBA' age, msize(small) mcolor(red)
, name(dd_age, replace) legend(off)
title("Probability of voting PM's party, diff btn men and women", size(vsmall))
xtitle("Age") ytitle("Diff prob men and women")
nodraw
xlabel(0(1)100,grid angle(45) labsize(vsmall))
ylabel(1(0.001)1,grid labsize(vsmall))

Even if just coding scatter `diffBA' age, I get nothing. Any advice?

[Nick Cox]

scatter wants to see a variable name in the position where you have put diffBA. If I understand your code correctly -- and your code example isn't reproducible -- diffBA will at best contain a numeric constant, so that won't work.

You seem to be asking for 100 graphs....

I think this is an example of the XY problem -- http://xyproblem.info/ -- and that you are re-creating what Stata will do for you willing and concisely, perhaps through predict and/or margins and marginsplot.