topic Re: How to quantitatively compare particle distributions?? in Discussions

How to quantitatively compare particle distributions??

seismicfish — Fri, 09 Jun 2023 00:41:12 GMT

Ok, so I have a large data set that has particle size distributions (surface area for each particle in the sample) for over 100 samples and I need a way to quantitatively (or semi-quantitatively) compare them to the whole data set or sort into groups with similar distribution characteristics. The vast majority of the points are in the very small size fraction but due to the very small size do not contribute proportionally all that much to the total area.

I have attached sample plots of three representative samples with visually different distributions.

-Histograms of the raw data where the y axis is the particle size

-xy plots of the particle size (bin=x axis) vs the percentage that size bin contributes to the total area (y axis). Two versions are shown, linear and log x axes (obviously log normal distributions here).

Eventually, these need to be visually comparable to one another (classes?) when the samples are plotted in real space relative to their collection point to delineate spatial patterns (purely visual). I have been banging my head on this matter for weeks. If anyone could help me I would be greatly obliged!

Thank you!

Re: How to quantitatively compare particle distributions??

peng_liu — Thu, 28 Oct 2021 20:35:56 GMT

How about comparing non-parametric distribution estimates of individual samples? Here is an example using the attached simulated data.

The report also comes with tests to see whether they are from the same distribution.

See Documentation: Life Distribution - Compare Groups

Re: How to quantitatively compare particle distributions??

ron_horne — Thu, 28 Oct 2021 20:42:21 GMT

Hi @seismicfish ,

for starters i would stack the data, then you should have one column with the particle size and the other with the location (GS31, Gs32 and GS35). with that you can use some of the options in the Fit Y by X platform including densities and other visual options. Otherwise, the stacked data is also more useful in the graph builder.

send us a data sample if you want some inspiration.

ron

Re: How to quantitatively compare particle distributions??

seismicfish — Thu, 28 Oct 2021 21:00:46 GMT

Sounds good as I am not quite sure what you are explaining. Here are some data to play with. Thanks!

Re: How to quantitatively compare particle distributions??

ron_horne — Thu, 28 Oct 2021 21:33:41 GMT

Hi @seismicfish,

Please open the attached files and let us know if any of this is useful?

Re: How to quantitatively compare particle distributions??

seismicfish — Fri, 29 Oct 2021 13:16:19 GMT

These could be very helpful. Thank you. How did you translate all the data into a single column for both files though?

Re: How to quantitatively compare particle distributions??

Mark_Bailey — Fri, 29 Oct 2021 13:37:26 GMT

I agree with @peng_liu about using Life Distribution to explore and test these distributions. I also recommend the Functional Data Explorer. This platform is for functional data, such as yours. You can learn more about it here.

Re: How to quantitatively compare particle distributions??

seismicfish — Fri, 29 Oct 2021 15:21:08 GMT

Dang. The FDE looks like it may be the ticket but unfortunately I don't have JMP Pro, just standard. Thanks though.

Re: How to quantitatively compare particle distributions??

seismicfish — Fri, 29 Oct 2021 16:36:13 GMT

This is great, peng_liu. I finally managed to replicate your script with my (monstrous) data set, but how do I get the group homogeneity tests? Thanks!

Re: How to quantitatively compare particle distributions??

peng_liu — Fri, 29 Oct 2021 18:08:55 GMT

It is not far below the main plot. The outline node is closed by default. You need to open it to see the content.