I want to know if anybody uses a tool to autogenerate a distribution that fits the sample data. Eg: I have a dataset of say 100 values. I want to just provide the tool the 100 values and it spits out the right kind of distribution that fits the data well (Eg: normal/binomial/uniform).
I want to automate the process of generating such distributions instead of manually figuring out what distribution to pick.
I already found my answer. FlexSim has a tool called "ExpertFit" that does this seemlessly.
Please understand that if we can do this reliably then statistics is doomed assuming I understand the question... Good luck, David Booth
Reinforcing comments from David Eugene Booth above, it is next to impossible to know for sure what is the true distribution based on a sample of, say, 100. There are dozens of distributions that cannot be differentiated from a sample of Normal, or even binomial, data. So generating an artificial data set requires that you make several assumptions (shape, rate, dispersion, intercorrelation, etc.), hopefully based on theory and/or domain knowledge.
An engineering simulator, such as FlexSim, probably has options for generating distributions that are binomial, triangular and normal. These are deamed "close enough" for most simulation purposes, because they're easy to communicate, and distributions are the least important influences on bottlenecks or vicious cycles in modelling a complex system.
If you need something more sophisticated, then I'd suggest the SimStudy package for the R Statistical Programming Language. There are probably other R packages that do good job too.
https://cran.r-project.org/web/packages/simstudy/vignettes/simstudy.html
- Badges
- Science topic