At the pixel level, which of these approaches is a more accurate depiction?

I suppose that it is pretty intuitive that an alpha of 0.5 would be somewhat similar to a "screendoored" object. But in a similar manner, would undersampling - ie taking an alpha = 1 yellow triangle and sampling every other pixel - result in the same?