Is it a good idea to use PERT analysis when planning a software project?

A friend recently suggested I use PERT analysis when estimating the time it will take me to complete projects. Put simply, PERT analysis is a technique that depends on your optimistic estimate, your pessimistic time estimate, and your normal time estimate. You then use a formula to calculate the expected time estimate. Here's an example: Activity: Develop module B Predecessor: Learn to use the necessary APIs Optimistic time estimate (O):3 hours Normal time estimate (M):5 hours Pessimistic time estimate (P):9 hours Expected time: (O +4M + P) ÷6 = (3 +5*4 +9) /6 =5.333 hours You do that for each task in the project, and then draw a gantt chart with the expected durations. It's supposedly accurate for giving an estimate for large projects. I was wondering, have any of you used it while planning software projects and would you recommend it?