>>14283727>but is this a good approach??It's definitely the most common approach, especially for something as simple as the task you are describing. It might not be a good approach in some extremely rare, esoteric example requiring state of the art multi-variate non-parametric Bayesian analysis - which isn't the problem you're facing.
>i was told that this is not a good way. That curve fitting to an input distribution is bad.thats a very subjective take. The whole point of the QQ plot and the Chi-Square test is to confirm that it is in fact a good distribution fit.
Common distributions such as the ones you posted are "common" and "well-known" for the reason that different parameterization can accommodate a large profile of different distributional shapes.
> i have an input distribution of how long one >tasks takes to finish.The goal is to use a >montecarlo Simulation to predict how long a >certian number of task would take to >finish.While only using the input distribution >you have the error that any task can not lie >outside of the distribution.Did they give you raw samples of data from the input distribution? If so, I don't see why curve fitting the input distribution is bad, unless you are doing something foolish like choosing a distribution that has a different support than the input distribution or is discrete/continuous when the input distribution is not.
Sorry for the formatting, I'm drunk and retarded and on my phone.
