Goodness of Fit - significance

[Lamya96]

Hi,

I'm using the distribution analysis tool to fit the distribution to the data,  however, I'm not able to understand the goodness of fit results, the significance numbers are not clear to me with the > and < sings. How can I translate the significance in the pic below?

In the pic below, does this mean that based on KS test that the normal and lognormal are the best fit to the data? since the significance numbers are higher than 0.15 (>0.15), this means they are higher than 0.05, hence, they fit?

Any kind of help will be appreciated.

Thanks,

[CharlieS]

Hi @Lamya96 

The Distribution Analysis tool is very powerful in that it offers multiple tests for multiple distribution types, but each test has strengths and weaknesses depending on the situation. The first step we need to take here is to understand the nature of the data being testing. What can you tell us about your data (number of observations, value range, sampling methods, etc.) ? Sharing the complete output from the Distribution Analysis tool would also be helpful. 

It's been a minute, but I believe the Kolmogorov-Smirnov results are only provided for the Normal and Log distributions because Gamma and Weibull distributions are not tested effectively without special modifications to the data or testing method. My Google results on the subject were mostly ways of how to force these tests to work in specific applications. 

[Garabujo7]

Hello @Lamya96 ,

Adding to the conversation, you may use this:

From this articlehttps://www.thedataschool.co.uk/andrew-lehm/alteryx-stats-tools-for-beginners-part-2-data-investigat...  to better understand goodness of fit.

Hope that helps

Gabriel