Plot margins for odds ratio

[David HU]

Dear stata users,

I recently got some difficulties switching from the command "gam" to "margins"

What I want is to plot the odds ratio and 95% CIs from a general additive model, starting at OR=1 for the smallest X value (X is continuous). Here is some of my old codes:

The arbitrary value "0.06043" is the fitted GAM_mu value for the smallest serd value.
First of all, is my GAM code correct? I assume the fitted GAM_mu is the odds, right?

gam stroke X, family(binomial) df(6)
gen strokeor= GAM_mu/0.06043
gen strokeorhigh=strokeor+strokeor* e_X*1.96
gen strokeorlow=strokeor-strokeor* e_X*1.96

scatter strokeorhigh X if X<=10, graphregion(fcolor(white) lcolor(white)) connect(l) clpattern(dash) msymbol(i) ///
|| scatter strokeor X if X<=10, graphregion(fcolor(white) lcolor(white))connect(l) msymbol(i) ///
|| scatter strokeorlow Xif X<=10, graphregion(fcolor(white) lcolor(white))connect(l) clpattern(dash) msymbol(i) ///
ytitle("OR for Stroke") xtitle("X (umol/L)") legend(off) scheme(s2mono)



Now I'd like to control some additional variables and use the "margin" command, but I have some difficulties to translate the code above.
The marginsplot or marginscontplot always plot the probability, not the odds ratio. How can I get do that?

First just assumes a linear relationship:

logit stroke X age sex married college income job sbp smoking bmi pinact epa dha

What should I do to get a plot for Odds ratio for stroke, start from 1

Later on, I would like to add test the nonlinear trend with the "rc_spline" command, What should I do? I tried something like this, but it does not work.

rc_spline X
logit stroke _SX* age sex married college income job sbp smoking bmi pinact epa dha

predict logodds, xb
predict stderr, stdp
gen odds=exp(logodds)
gen odds_lb=odds-1.98*exp(stderr)
gen odds_ub=odds+1.98*exp(stderr)


Thanks a lot

[David HU]

I revised my codes as follows:

logit mi c.X##c.X##c.X age sex married college income job sbp smoking bmi pinact epa dha, or
margins, at ((means) age sex married college income job sbp smoking bmi pinact epa dha X=(-0.5(0.1)10)) expression(exp(predict(xb))/0.09156)
marginsplot, recast(line) recastci(rarea) scheme(s2mono) xdim(X)

The Arbitrary value 0.09156 is the margin when X==-0.5

Does that make sense?