Which of the following is an appropriate confidence interval to indicate significance for an OR=4.5? A. 2.6 to 4.4 B. 1.3 to 6.7 C. 0.5 to 6.7 D. -3.4 to 4.5

Answer 1

Your question was:

Which of the following is an appropriate confidence interval to indicate significance for an OR=4.5? A. 2.6 to 4.4 B. 1.3 to 6.7 C. 0.5 to 6.7 D. -3.4 to 4.5

A way to think of this is, odds = how likely is this to happen to an individual if they are positive for the Independent variable; and ratio = comparing two things.

If your OR is 4.5 times more likely to happen with a positive-possession of the independent variable than those without it, then you would have a positive (e.g., it's happening) correlation between the IV and the DV.

So, an actual positive association would never "straddle" one, so you can throw out the 6.7 - 0.5 immediately (a "straddle" is when one number is greater than 1 and the other less than 1, with 1 between them. This would indicate you "don't know" if this is a risk or protective association, i.e., you don't know if they are 6.7 times more likely to get disease X or protected [0.5, or 50% less likely to get disease X]. Therefore, a statistic that "straddles one" is inconclusive, i.e., doesn't tell you anything and you can't use it to say "anything" is happening).

Then there is the negative number, which shouldn't be possible with an odds ratio. So that's out.

And finally (because we only need to eliminate one more set of CI's), we can throw out the CI's that don't include the "finding", i.e., 4.5. That would be the 2.6 - 4.4 (throw it out).

Why? Since the CI is the "Blurr" in the "vision" of the equation, the computer may give you a specific number, but can only be confident that it is "Somewhere" inside the CI.

To think of it in analogy: you are looking at a video of a crime. The bank robber runs out of the bank, and you can see their height next to the door stripes (used by the FBI to tell how tall a robber is), but the graininess of the film is too blurry to see if they are exactly 6'2, as the pixals are too large on the screen (all you see is blurry blocks of color when you look really close). So, you can say, well they are definitely taller than 5'10" tall and shorter than 6'4" tall. So, your confidence interval is between 5'10" and 6'4", with your best "guess" at 6'2" tall.

See? You wouldn't say, well the Confidence Interval is between 5'10" and 6'4", with our best guess at his/her height at 7'0", because that's outside of the blurry blocks of color! We KNOW s/he is shorter than 6'4" (or equal to), so their height HAS TO BE INSIDE OF THE CI!

All of your Epidemiology statistics will have your "best guess" (i.e., the answer the computer spat out) between the CIs. All of them!

So that leaves: CI of 1.3 - 6.7 (ranging from a little more likely [30%] for the IV to be associated with getting the DV, and 670% more likely to get disease X if one is positive for the Independent Variable. Your interpretive summary of this statistic should include verbiage such as the previous sentence.

I hope this helps. Message me if you are still confused. :)

Answer 2

Significance if the value lies outside the confidence interval (A).