Why is it hard to prove about the Fisher's factorization theorem in case of continious?

45 Views Asked by At

I'm learning about the proof of Fisher's factorization theorem online. Almost all the proofs of it in case of continuous is using the one-to-one transformation. But I can't find the necessary of transformation.

I assume that $f_{\theta}(x)=h(x)g(T(x);\theta)$,then I want to prove $T(x)$ is sufficient for $\theta$

I think the proof:

$f_{\theta}(x|t)=\frac{f_{\theta}(x,t)}{f_{\theta}(t)}\\ =\frac{f_{\theta}(x)}{\int\ldots\int_{x;T(x)=t}f_{\theta}(x,t)dx}\\=\frac{h(x)g(t;\theta)}{\int \ldots \int_{x;T(x)=t} h(x)g(t;\theta)dx}\\=\frac{h(x)g(t;\theta)}{g(t;\theta)\int \ldots \int_{x;T(x)=t}h(x)dx}\\=\frac{h(x)}{\int\ldots\int_{x;T(x)=t}h(x)dx}$

The last formula is independent of $\theta$, so the conditional pdf given T(x)=t is sufficient statistics.

Is there anything wrong with what I wrote ?

Thank you for helping me and please bear with my poor English!