How long do you keep identifying the distribution of visual data
Yukio Takahashi Representative Director, BioStat Research Institute, Ltd.
Identifying what distribution the obtained data applies to is a first step in statistical analysis.Many methods have been provided for visually confirming the distribution of data using JMP's "univariate distribution".Traditionally, judgment was made mainly on whether fitting of a normal distribution is possible or not.For JMP's "univariate distribution", the fitting of the probability density function to the histogram was also supported previously.When "All" is selected in "Fitting continuous distribution" in the latest version, many distributions are listed in order of -2 times logarithmic likelihood and AICc (corrected Akaike information criterion) in order of magnitude, and for distribution fitting AICc Can be determined.When ranking of distribution fitting using AICc was evaluated with various data, it was meaningful.Since log likelihood is being outputted, it seems that maximum likelihood method is used, so we used Excel to perform additional test of JMP calculation method.Using AICc, it seems like a good opportunity to break away from the conventional visual ambiguity judgment when investigating what kind of distribution fitting is appropriate.Since the same function is also in "univariate of lifetime", we also want to discuss new distribution fitting.
