CALCULATION & LISTING OF DISTRIBUTION FEATURES OF A VARIABLE

[#1]

For a classification study I need to calculate the distribution features (distributions T & NON T are different) of PC Scores in attached data set. A JMP file has to be constructed where per participant (1 to 16) and per classification T/NON T) and per sample of 100 rows the following distribution features have to be calculated and listed: mean | standard deviation | variance | skewness | kurtosis | percentiles 10th/25th/50th/75th/90th | entropy 

How to do this?

[#2]

You can for example use Distribution with By columns

Then add Customize Summarize statistics and Quantiles as you need to

After that, you can right click on the tables in the platform and create Combined Data Table

Then you can combine those two tables with Concatenate (might require some column renames)

[#3]

Sorry for delay, thanks I am going to try this asap

[#4]

Hi @frankderuyck,

Besides the excellent suggestion proposed by @jthi, you could also use the Tabulate platform that enables you to calculate a lot of statistics from your data table. Some other statistics, like skewness and kurtosis, can be easily calculated with column formula from basic statistics like means and std dev computed with tabulate: https://www.itl.nist.gov/div898/handbook/eda/section3/eda35b.htm

Hope this second option may also be of interest,

[#5]

Thanks Victor & @jthi, for you useful suggestions, however thinkk there is a misunderstanding so I need to clarify more clearly my task. I need to make a classification model for two movement types T and NON T. As shown in graph below the PC Score distributions of T & NON T (cfr. file above in attachment) are different so I need to find a model based on the score distribution characteristics. To do so, for all 16 participants I want to draw samples of 100 consecutive rows from each movement type and report in rows of a new table the distribution properties so I can generate a ML binary classification model T/NON T. 

Hope this is sufficiently clear; how to create a table with in rows, for all participants, the T/NON T distribution properties of 100 row samples? 

[#6]

Create a formula which lets you select rows you are interested in (you might have to add your classification column to the formula as grouping column)

Col Rank(:Participant, :Participant) <= 100

create subset/exclude not wanted rows and then calculate the statistics.

[#7]

Hi Jarmo, does not work, 100 first rows get a label 1, rows > 100 are labeled 0 ?

[#8]

you might have to add your classification column to the formula as grouping column

---

Col Rank(column, <ByVar, ...>, <<tie("average"|"arbitrary"|"row"|"minimum")) 

There are also plenty of other formulas you could use