bootstrapped CIs for standardized indirect effects using sem

[Daniel Maar]

Dear all,

I struggle with calculating the bootstrapped percentile confidence intervals for a standardized indirect effect after running sem in Stata 14. So far, I only managed to get bootstrapped percentile CIs for the unstandardized indirect effect with the following code:

capture program drop bootmm
program bootmm, rclass

sem (X -> x1 x2 x3) ///
(M1 -> m11 m12 m13 m14) ///
(M2 -> m21 m22 m23 m24) ///
(Y -> y1 y2 y3 y4), vce(sbentler) cov(e.M1*e.M2) standardized

return scalar ind1 = (_b[M1:X] * _b[Y:M1])
end

bootstrap r(ind1), reps(2000): bootmm
estat boot, percentile

I tried to get the confidence intervals for the standardized effect by modifying one line in the code (highlighted in bold):

capture program drop bootmm
program bootmm, rclass

sem (X -> x1 x2 x3) ///
(M1 -> m11 m12 m13 m14) ///
(M2 -> m21 m22 m23 m24) ///
(Y -> y1 y2 y3 y4), vce(sbentler) cov(e.M1*e.M2) standardized

estat stdize: return scalar ind1 = (_b[M1:X] * _b[Y:M1])
end

bootstrap r(ind1), reps(2000): bootmm
estat boot, percentile

However, it seems that the combined use of estat stdize and return is not allowed.

Thank you in advance for your valuable help!
Daniel