to snoo-finity ... and beyond!
to snoo-finity ... and beyond!
Knowing the LR+ value = 10 does not help to solve this Q because estimating where "10" should fall on an axis is arbitrary. Also, the data points are coordinates -- they have an X-value and a Y-value (X, Y).
The way to approach this Q is to know that a high specificity means that a positive result is very, very likely to be a true positive.
Suppose that the specificity is 0.99 -- this is 99% specificity. Then, you look at the graph. The X-axis is "1-specificity." So, suppose the best test has a specificity of 99%. Then, calculate 1-specificity = 1 - 0.99 = 0.10. You would then chose the datapoint that corresponds to having an "X-value" that is closest to the origin. In this problem, it corresponds to data point "A."
You don't even need to know a specific specificity value to solve this problem. All you need to do is understand that if the specificity is extremely high, you will need to find a datapoint that is closest to the origin -- at least for the value in the X-axis in the coordinate of the data point -- because the X-axis corresponds to a calculation of "1-specificity".
The words "most effectively" confused me. I thought to myself "even if it has the highest specificity, it's not very effective if it's got super low sensitivity -- since it will miss a lot of the true +ve's (failing to rule in cancer for the ones that get missed)." That was a story I told myself about their use of the words "most effectively" lol.
This is a likelihood ratio. LR+= Sens/1-Specif
Any value greater than 10 (per first aid) indicated "usefulness of diagnostic test" which is comparable to PPV (ruling in a dz). Point "A" is the closest mark to where 10 should be on the Y axis.
submitted by brise(18), 2019-06-02T23:59:26Z
The question is asking what point would be the most likely to rule in cancer, and high specificity when positive rules in cancer. The highest specificity value is A, bc the the X axis shows (1-specificity)!