the graph looks great! Im new to R and I do have some problems to get the Syntax work. Could you help me? I’ve got a data frame for which I need all corrrelations in a scatter plot (23 Variables, I know its a lot). The data is ordinal with 4 scale steps (1 to 4). Can I read in the wholde dataset or should I list all the variable names in? I deleted all other Variables in the dataset, so there are only the 23 Variables with no missings.
Thanks a lot!

If any reader wishes to offer a fix – I’d be glad to incorporate it.

In such a case, I would suggest:
1) using bootstrap/permutation methods in order to calculate the P values.
2) Checking what methods exists for this case in the literature and see if they’ve got an R implementation.

Cheers,
Tal

First of all, thanks for putting together the code for Likert-scale correlation matrix; it has helped me a lot.

My question is about the use of Spearman’s rho: as you know, Spearman’s formula cannot calculate p-values with good precision, when there are ties in the ranks. Such ties usually exist in survey data. Do you suggest using Pearson’s corr. coeff. instead?

I’d really appreciate if you could comment on this.

Also thanks for pointing that the graphic doesn’t use solution 2 (I’ll correct that in the article).

The reason I wrote it is that the code give an hidden option to use solution 2 (I just didn’t feel it added enough in this case to include it).

For people who will want to implement solution 2, please remove the “note #” sign from:
#the.col < - heat_hcl(length(x))[z]
And
inches=F, bg= “grey”,#the.col ,
In the:
panel.smooth.ordered.categorical function.

I am sure it could be improved more. My biggest problem (with both the colors and the sizes), is how to make sure they scale “well” for various situations in the data.
If you have any thoughts – be welcomed to share them!

Cheers
Tal

Small typo: the graphic uses solutions 3-4, not 2-4 as the dots have no color gradient.