Maybe it's just me, but I found Integrals more intuitive as a continuous "sum" of probabilities in a distribution. So, if you want to marginalize a discrete dis

Maybe it's just me, but I found Integrals more intuitive as a continuous "sum" of probabilities in a distribution. So, if you want to marginalize a discrete distribution, you just add all its components over the conditional variable. If you want to marginalize a continuous distribution, you integrate. I guess you could say that you're simply enumerating infinitesimal densities?