Creating a scatterplot matrix with J

Question

I am doing some data science on the commandline and discovered J as handy tool for basic analysis (statistics) tasks. After gaining some initial feeling for the language some years ago, I have to restart as total noob. It would be great to get some help for where I am stuck now.

I have a set of measured attribute and want to display the correlation visually. The view is often called scatterplot matrix, a good example is here. I would think a similar figure can be created in J using its great plot library.

What I have done so far:

NB. For sake of brevity, my test data is simulated here comprising 4 attributes with 7 measurements:
   ]A=:i. 7 4
 0  1  2  3
 4  5  6  7
 8  9 10 11
12 13 14 15
16 17 18 19
20 21 22 23
24 25 26 27

NB. With the aid of a sparse matrix I created the index matrix of the multiplot matrix/layout
   ]Idx=: 4 $. $. 4 4$1
0 0
0 1
0 2
0 3
1 0
1 1
1 2
1 3
2 0
2 1
2 2
2 3
3 0
3 1
3 2
3 3

NB. Creating a single plot work fine with this sentence, here showing attribute 1 vs 3.
'TYPE marker' plot 1 1 <;.1 |: 3 1 {"1 A

NB. And here is my failed approach for creating a (multiplot) scatterplot matrix
   pd 'multi 4 4'
   3 : 0''
for_ijk. Idx do.
pd 1 1 <;.1 |: ijk {"1 A
end.
)
   pd 'show'

First of all, doing this via a loop feels wrong in J. Is there a better way to inject the index for each plot element?

Second, the loop does not correctly feed data to plot (pd) but I could not find a way to append all the results to big matrix and feed only this to pd.

Any ideas are highly welcome! Many thanks in advance.

Answer 1

Your approach is fine, except that pd-multi needs, as far as I know, all data at once as a boxed list of 2-elements-lists:

(options1;data1) ; (options2;data2) ; ... 

One possible way to do this is:

options =: 'type marker'
pd'reset'
pd'multi 4 4'
pd (options; <)"1 (A&({"1~)) &.> Idx
pd 'show'

assuming options is common for all plots.