(It's many, many years since my last foray into statistics so please be gentle..)

In its simplest form my question is this: I have developed a model using regression analysis that provides a good fit to my data, using multiple input variables. Given all the information I have from the regression model, given two sets of independent variables to use as inputs, how do I calculate the probabililty of one real observation being greater than another real observation?

I searched the forum and the internet, and the closest I have to an answer is along the lines of (from this forum):

"

Compute the difference distribution and determine the probability that it has a value of < 0.

mean of difference distribution = mean1 - mean2

std dev of difference distribution = sqrt(var1 + var2)

note--> var = (std dev)^2

"

however, I'm not sure how to implement this, as it implies that I have data about two distributions, but the regression analysis I have done only seeks to fit one distribution, or am I confused...?

Please help; I think what is underlying my question is a lack of understanding around some basic concepts, so I would be keen to learn as much as possible and correct these.

Cheers

Tim