Number stresser node

Hi,

I was wondering if someone understands how the stresser node works. The description is not very explicit when it comes to number columns. 

Lets assume I just chose one number column,

Percentage to stress per column is set to 1, so I always select that column for selected row

Percentage to stress per row is set to 1, so I select each and every row in the data set

What is the difference between in range and out of range stressing? 

As far as I can see, after testing it, in range stressing is just swapping a value in a given cell with a value drawn from uniform distribution over minimum - maximum range in that columns?

What distribution does the out of range stressing use? 

Also, I understand, that percentage per row is probability that row is selected, and percentage per column is conditional probability that column is selected given row, so it's just like flipping a loaded coin, and doesn't affect the magnitude of the noise that's added to the values? 

Any help will be appreciated. 

Hi Joseph,

I understand it (I wrote it :D) 

The stressing methods are the following: consider your minimum and max of the column values is 0-1.

Than the in-range methods will append a random number between 0-1

The out-range method will append a random number outside 0-1, this is is done using a gaussian distribution starting from min/max and using the range as standard distribution.

Also, I understand, that percentage per row is probability that row is selected, and percentage per column is conditional probability that column is selected given row, so it's just like flipping a loaded coin, and doesn't affect the magnitude of the noise that's added to the values? 

Yes this is correct. The percentage to stress per column only  is used if you have more than one column selected.

What are you doing with the node?

Cheers, Iris 

Thank you for the explanation Iris. 

I was thinking about using it for stress testing costs in my model, but unfortunately it doesn't offer a lot of control over how the stressing is done. I think I'll stick with permutations and math formulas, but it's nice to have a node like that, in case I need quick solution in the future.

Thank you,