I've been using the RiskAmp plug-in to Excel 2003 SP3 for a while now, and was recently looking at the PDFs for the input functions PERTValue, BetaValue, and LogarithmicValue. These all seem to show a spike in the resulting PDF curves, that I can't account for, and won't go away with higher numbers of iterations.
Can someone help me understand why this is so, please? Happy to share the Excel spreadsheet if needed.
In case it has a bearing, I'm using latin hypercube sampling (can't imagine why I'd ever *not* want to use this option, unless it's for performance reasons?). This seems to make all the PDFs converge more quickly than not using LHS.
Can you send me a spreadsheet to work from, or in the alternative, just post some specific model parameters? I'd like to understand what you're seeing.