Assess statistical significance of two PPVs

#1
Assess significance of the difference between two PPVs

I recently validated a medical diagnosis in order to compose a study population. To do so, we classified each patient with a diagnosis of interest as probable, possible, or unlikely case, depending on how likely we believed the diagnosis to actually be true.

I then calculated two different PPVs, one probable vs. unlikely and the other possible vs. unlikely, and I would now like to test if the resulting PPVs are statistically significantly different from each other. The numbers are indicated below (++ true positives, +- false positives, -+ false negatives, -- true negatives):

Probable vs unlikely: ++45 +-6 -+0 --11 PPV 0.88 (95% CI 0.79-0.97)

Possible vs unlikely: ++48 +-8 -+0 --11 PPV 0.86 (95% CI 0.77-0.95)

What test do I have to use to see if the difference between the two PPVs is statistically significant?

Thanks for your help. I hope the information above is sufficient.
 
Last edited:

hlsmith

Less is more. Stay pure. Stay poor.
#2
First, you may want to reevaluate level of significance value, in that you have multiple comparisons. This is given you have not already done this.
 
#3
Thank you for your response. Sadly, I do not understand what you mean by 'reevaluate level of significance value, in that you have multiple comparisons'.

On a further note: From the things I have read so far it seems like the Pearson chi-square test might be a possible way to assess the equality of the two PPVs. Is this possible?
 
Last edited: