Task Percentiles in Pertmaster

Matthew, I'll have to check my archive If I have any details about this specific warning, but from memory - When you run the analysis in Pertmaster/PRA, you need to specify in the options that you want certain percentile dates calculated and stored (eg. P50 Start Date, P50 Finish Date). If you've then changed any of the inputs of the risk model, this renders those calculated dates as invalid - they may not represent the Start/Finish dates of the revised model, and you need to run the analysis again, with the option to calculate the risk percentiles on again.

For what its worth - the more modern risk analysis tools, automatically calculate these dates and do it significantly faster than Pertmaster/PRA does.

Thanks a lot for your advice. I have tried to run the analysis twice, and i got the same warning message. And my setting on risk percentiles is shown below, I believe I didn't set something strange there......

Thanks a lot for your advice. I have tried to run the analysis twice, and i got the same warning message. And my setting on risk percentiles is shown below, I believe I didn't set something strange there......

the warning message pop up after the analysis finish running all the iterations.

Matthew,

I can't find anythign in my archives, and I no longer have access to Oracle Support for Primavera Risk Analysis. When do you get this message - what is the action you're performing? Are you getting P5 and P95 values displayed in the columns that match the results you see in the Distribution Graph?

Matthew,

Thanks - I can't even re-create this warning so cannot understand why it occurs. Does the warning come up if you turn off the option to calculate risk percentiles?

i just found that the only way to get rid of the warning is to run under 1000 iterations...

Matthew,

Interesting! Do you notice any difference in the results? I shouldn't expect any.

surely there was some differences between the case of 500 iterations and 5000 iterations, especially towards P95/P100. 

Matthew, thats hard to asses without knowing more about your model. If you have an number of risks that are in the 0-5% probability range, then certainly, it might make a small difference.