A resposta curta é "sim, você pode" - mas você deve comparar as Estimativas de Máxima Verossimilhança (MLEs) do "grande modelo" com todas as co-variáveis nos dois modelos ajustados a ambos.
Esta é uma maneira "quase formal" de obter que a teoria das probabilidades responda à sua pergunta
No exemplo, e Y 2 são o mesmo tipo de variáveis (frações / porcentagens), portanto são comparáveis. Assumirei que você encaixa o mesmo modelo nos dois. Portanto, temos dois modelos:Y1Y2
l o g ( p 1 i
M1:Y1i∼Bin(n1i,p1i)
M2:Y2i∼Bin(n2i,p2i)log(p 2 ilog(p1i1−p1i)=α1+β1Xi
M2:Y2i∼Bin(n2i,p2i)
log(p2i1−p2i)=α2+β2Xi
Então você tem a hipótese que deseja avaliar:
H0:β1>β2
{Y1i,Y2i,Xi}ni=1
P=Pr(H0|{Y1i,Y2i,Xi}ni=1,I)
H0
P=∫∞−∞∫∞−∞∫∞−∞∫∞−∞Pr(H0,α1,α2,β1,β2|{Y1i,Y2i,Xi}ni=1,I)dα1dα2dβ1dβ2
A hipótese simplesmente restringe o alcance da integração, portanto, temos:
P=∫∞−∞∫∞β2∫∞−∞∫∞−∞Pr(α1,α2,β1,β2|{Y1i,Y2i,Xi}ni=1,I)dα1dα2dβ1dβ2
Como a probabilidade depende dos dados, ela será fatorada nos dois posteriores separados para cada modelo
Pr(α1,β1|{Y1i,Xi,Y2i}ni=1,I)Pr(α2,β2|{Y2i,Xi,Y1i}ni=1,I)
Now because there is no direct links between Y1i and α2,β2, only indirect links through Xi, which is known, it will drop out of the conditioning in the second posterior. same for Y2i in the first posterior.
From standard logistic regression theory, and assuming uniform prior probabilities, the posterior for the parameters is approximately bi-variate normal with mean equal to the MLEs, and variance equal to the information matrix, denoted by V1 and V2 - which do not depend on the parameters, only the MLEs. so you have straight-forward normal integrals with known variance matrix. αj marginalises out with no contribution (as would any other "common variable") and we are left with the usual result (I can post the details of the derivation if you want, but its pretty "standard" stuff):
P=Φ(β^2,MLE−β^1,MLEV1:β,β+V2:β,β−−−−−−−−−−−√)
Where Φ() is just the standard normal CDF. This is the usual comparison of normal means test. But note that this approach requires the use of the same set of regression variables in each. In the multivariate case with many predictors, if you have different regression variables, the integrals will become effectively equal to the above test, but from the MLEs of the two betas from the "big model" which includes all covariates from both models.