superSubset question please

[Sarah Wang]

Hi everyone, 

I am wondering if there is any commomly accepted parameters when we use superSubset () for the analysis of necessity. 

superSubset(data = mydata,
                      outcome = ,
                      incl.cut = ?,
                      cov.cut =?,
                      ron.cut =? )

I understand that 0.9 is the lowest for incl.cut from literature. How about cov.cut and ron.cut? 

Thank you

Kind regards

Sarah 

[Adrian Dușa]

I am not sure there are specific (exact) values for these thresholds, it depends from case to case.

In their work about Intersectional Inequality, Ragin and Fiss used a thresholds of 0.745 for one of their analysis of sufficiency. The reason they gave was due to a "substantial gap" that was noticed compared to the next configuration score in the truth table.

You also did not mention if your superSubset relation is about necessity of sufficiency, I assume necessity given your ron.cut question.

Generally, the consistency for the sufficiency should be located around 0.8 or more, but not less than 0.7

For necessity, the rule should be about the same but the literature is usually infused with lower consistency scores compared to the sufficiency.

Do please note that coverage for sufficiency is the same thing as consistency for necessity.

The relevance ron.cut is only used for necessity, and should probably be no less than 0.6 but all these thresholds should never be taken mechanically. A visual inspection of the XY plot is always the best advice, function of which the researcher more accurately inspects:

- which particular cases are located in which quadrant

- which cases seem to be deviant (and further research why)

- revisit the calibration perhaps, or even the initial raw measurement

- begin the process once again, and why not perhaps the deviant cases (and the consistency scores) get improved because of this dialog.

I hope this helps,

Adrian