Hi,
I am trying to identify a distribution model for my real data, I used Pareto family distributions, Generalized pareto and Pearson family distributions. However I am still stuck of the goodness of fit test. When I use chi-squate test, it gives a p-value equal or close to 1 (maybe I am not doing the test right). Is there a good "goodness of fit" test that can work on theese distributions.
Thank you,
Saad,
I would suggest you compare your data (raw or modelled/adjusted) to the theoretical distribution via a graphic visualization. I do this extensively in my work, using a user-writtten program in Stata called hangroot (written by Maarten Buis and available for download from within Stata by typing ssc install hangroot).
I hope this helps
Ariel
Saad,
To check goodness of fit I use extensively the R function 'qqplot'. This R function plots quantiles of both the model and the actual data on the same graph, if they lie over the 45-deg straight line the model is perfect. It is an easy, quick way to check different models for your data.
Once you have examined a number of functions to fit your data you can implement a quantitative measure of goodness of fit as described in: Ricci V. (2005). Fitting distributions with R (https://cran.r-project.org/doc/contrib/Ricci-distributions-en.pdf)
Hope this helps.
Augusto
- Badges
- Science topic
- Similar topics
- Statistics
- Survey