What inspired this Wish List Request? A user reported that in JMP18.1, we need to specify the control group (i.e. denominator) for "Sample X categoriy in numerator" in Relative Risk. This is very confusing and misleading.
Please describe the new or improved feature or solution you would like to see.
In the current JMP's Contingency platform, we have two parts to calculate risk differences and risk ratios. But they are not consistent.
For risk ratios, you can calculate them by Relative Risk and Equivalence Tests Command.
/* NOTE: The results by the following Relative Risk commmand are different between JMP17 and JMP18 */
dt = New Table("Test",
New Column("X", Character, Values({"A", "A", "B", "B"})),
New Column("Y", Nominal, Values([0,1,0,1])),
New Column("Freq", Values([10, 90, 50, 50]))
);
dt << Contingency(
Y( :Y ), X( :X ), Freq( :Freq ),
Contingency Table( 0 ), Mosaic Plot( 0 ), Tests( 0 ),
Equivalence Tests of Relative Risk(
1, 0.025, "Superiority Greater",
Response Group( 1 ),
Factor Group( "A" )
),
Relative Risk( 1, "1", "B" )
);
For risk differences, you can calculate them by Two Sample Test for Proportions and Equivalence Tests command.
dt = New Table("Test",
New Column("X", Character, Values({"A", "A", "B", "B"})),
New Column("Y", Nominal, Values([0,1,0,1])),
New Column("Freq", Values([10, 90, 50, 50]))
);
dt << Contingency(
Y( :Y ), X( :X ),
Freq( :Freq ),
Two Sample Test for Proportions( 1 ),
Equivalence Tests of Risk Difference(
0, 0.025, "Superiority Greater",
Response Group( 1 ),
Factor Group( "A" )
)
);
(a) In JMP18, we have four parts to specify the control group (or the treatment group) and the target response level for risk ratios and risk differences.
(1) The Control Level and Target Level column property.
(2) In the dialog called by Relative Risk command.
(3) Check boxes at the report for Two Sample Test for Proportions command.
(4) In the dialog called by Equivalence Tests.
But they use different terminologies for referring to the control level and the target level. I would like to recommend we should use "Control Level" and "Target Level" in all parts. Note that the "Control Level" is the denominator part, not numerator part, in the Relative Risk. And, the "Coltrol Level" is P0 in P1 - P0 in the risk differences.
(b) The different approximation methods are used for calculating confidece intervals (1)between Relative Risk and Equivalence Tests, and (2) between Two Sample Test for Proportions and Equivalence Tests.
I would like to suggest that old Relative Risk and Two Sample Test for Proportions also
calculate the confidence intervals (and the p-values) calculated by Equivalence Tests (so-called score confidence interval and score tests) becuase the score confidence intervals are consistent with the popular Pearson's chi-square tests.
(c) I would like to suggest that we should also support odds ratios in Euqivalence Tests, and add score confidence intervals for the old Odds Ratio commmand.
(d) I would like to suggest that Equivalence Tests should also support the Control Level and Target Level column property. I mean that the levels specified by these column properties should be selected at first in the dialog.
What would the value be to you and/or other users if this new feature is provided? Especially for pharamaceutical area, we should offer reliable statistical results.
Is this an external customer request? No.
Is there a competitive software package that has this feature or solution that we should be aware of? I don't know.
Will this impact a renewal, expansion, or adoption at a customer, and does this put an account at risk if we do not address this? I don't know.
