Never seen this done with chi-square, and would imagine that it quickly becomes very difficult when you have a project where changing task durations may redefine the critical path.

The more usual approach is through Monte-Carlo methods, about which you'll find plenty in other threads on this forum.

A.

I suggest you Google it, http://www.google.com.au/search?hl=en&q=chi+square+table&meta=

Chi squared tests are for discrete data sets (Go/No-Go) which I doubt would be appropriate for your project tasks as they are continuous data or outcomes (ie. time related).
As suggested Monte Carlo, (using software such as "Crystal Ball") would more likely be the way, analysing each tasks probability profile to give the probability of the overall project outcome.  

FOETS
The truth is out there. Anyone know the URL?

Thanks guys,  I will look into the Monte Carlo simulations.  The only reason I was looking into the Chi Squared table was because someone else had mentioned it to me.

K4rm4

Just food for thought, you might also want to check in to using a cumulative probability distribution (commonly known as an S-curve).  Basically it is a representation of the potential outcomes and their probability of occurrence.  You can use Excel and plug in the increasing probability on the vertical axis and the resulting outcome along the horizontal axis.

One tip, when using S-curves, it is helpful to overlay cumulative probability curves from the various alternative strategies or probabilities being considered in order to compare the various trade-offs of the curves.  

It doesn't handle near as many permutations as Monte Carlo, but doesn't sound like you need anything real complicated.  Monte Carlo simulation is a way of generating probabilistic results so you basically look only at risks and not neccesarily sequential decisions.  

Again based on your description, the probability of finishing a project on time, I would go with an S-curve.

Greg Lamberson, BS, MBA
Consultant - Upstream Energy
Website: www.oil-gas-consulting.com