topic Re: Probability plot for log-normal distribution fitting in Discussions

Probability plot for log-normal distribution fitting

mtrain — Fri, 09 Jun 2023 00:17:05 GMT

Hi,

Is there an option to view the probability plot when making a log-normal fit with a distribution analysis?

Thanks

Re: Probability plot for log-normal distribution fitting

HadleyMyers — Wed, 10 Jun 2020 08:53:04 GMT

Hi,

From the red hotspot of beside the name of your column in the Distribution platform choose "Continuous Fit -->Fit lognormal." Then, from red hotspot beside "Fitted Lognormal Distribution," choose "Diagnostic Plots." You then have the option of fitting the P-P and/or Q-Q plots. Is that what you're looking for?

Re: Probability plot for log-normal distribution fitting

gzmorgan0 — Wed, 10 Jun 2020 21:18:11 GMT

@mtrain,

@HadleyMyers provided you with a solution. However, if your lognormal distribution has a large mean and/or variance, the scaling on the QQ plot usually does not have the familiar custom probability plot that scales the Y axes like custom probability paper (yes I am showing my age, things we used pre graphical calculator/computer days).

I suggest you right click on the column of data you're fitting, select Transform and use Log, then Continuous Fit > Fit Normal.

Below is a screenshot of Analyze > Distribution > Continuous Fit > Fit Normal of 100 random lognormal(4,2) values transformed with Log. The table with script is attached.

Another suggestsion is to use the QQ plot that Hadley suggested and click on the Y axis of he QQ plot and change Linear to Log.

Re: Probability plot for log-normal distribution fitting

mtrain — Sun, 14 Jun 2020 23:26:40 GMT

Hi Hadley,
That is what I was looking for but without the confidence interval. But not a big deal.
Thanks for showing me around.

Re: Probability plot for log-normal distribution fitting

mtrain — Sun, 14 Jun 2020 23:31:16 GMT

Hi gzmorgan0,
I agree, making the log transformation is a great way to get the intervals on the plot too. I did this before but was also wondering about getting a plot for the log fit. The PP/QQ do a good job.
Thanks for the help.