Firstly, thank you for making the implementations opensource and also for the wiki on them.
mWEIGHTING.m when creating the linear weight matrix, the dist/maxdist division returns rounded-off int weight matrix for int categories. This leads to inflated scores. I think a datatype check could be incorporated to convert the categories or codes to double. This is just a corner case but it does not raise any flags unless inspected in depth. I think the same is valid for quadratic weighting as well.
jmgirard
commented
3 years ago
Firstly, thank you for making the implementations opensource and also for the wiki on them.
mWEIGHTING.mwhen creating the linear weight matrix, thedist/maxdistdivision returns rounded-off int weight matrix for int categories. This leads to inflated scores. I think a datatype check could be incorporated to convert the categories or codes to double. This is just a corner case but it does not raise any flags unless inspected in depth. I think the same is valid for quadratic weighting as well.