Thank you for the explanation. Ken Rockwell on his web seems to indicate that by downsampling in post processing, the noise get eliminated effectively the same as if you had a lower MP camera. So my question is:
1. Is his statement correct ?
His statement is essentially correct but he is not careful about his choice of words and you have changed the meaning.
We sometimes "downsample" to make images from two cameras the same size for a fair comparison. Downsampling doesn't "cancel" noise but it does concentrate the signal and noise from multiple pixels into one. Signal concentrates faster than noise so combining pixels always makes for a higher quality combined pixel. What gives the pixel it's quality is how much light originally went into it, concentrated or not.
2. By setting up camera at smaller JPEG size (ie 4256 x 2832), is it the same effect as if I was down sampling (resizing) in post processing in Photoshop or aperture.
No! Your smaller pixels will not be the same quality as pixels of the same size that were "concentrated" from a larger area. Remember you collected less light to make the smaller JPEG pixel than the amount of light that got squeezed into the downsampled pixel.