1
$\begingroup$

The material I am working with:

http://personal.vu.nl/a.f.de.vos/primer/primer.pdf

The example that I am working on can be found on page 2, see picture at bottom of post. It is stated that "Prior odds of 99:1 ...change by the result to posterior odds 99:13.7, a 12% probability." Later it says the posterior odds are 13.7:1, so 6.7% still rather different."

What am I missing here?

Could this not be described as decimal too:

.073 * .99 = .07227

In other words 7.2%, a number different than stated in the paper

enter image description here

enter image description here

EDIT: still having trouble with final part of the problem

now solving (1-x)/x = 0.072993, we arrive at x = 93.2%

and

so (1-x)/x = 7.226 or x = 1/8.226 = 12.16% probability

are throwing me off

enter image description here

$\endgroup$
  • $\begingroup$ note that an odds ratio less than one isn't commonly notated as a percent ... just like an odds ratio greater than one isn't likely to be expressed as a percent. if anything, an odds ratio might be followed by a multiply symbol, e.g. 0.07227× ... but that might get confusing when there's a variable $x$ lurking about. the $aa$:$bb$ notation is probably best for odds ratios $\endgroup$ – phdmba7of12 Sep 11 '18 at 22:17
1
+50
$\begingroup$

if the prior odds ratio is 99:1 (with a corresponding pre-felt probability of (1-x)/x = 99 ... x = 1% chance that the aunt has the ability)... multiplication by the likelihood results in posterior odds ratio of 99:13.7 = 7.226

odds are ratios of probabilities that add to unity

so (1-x)/x = 7.226 or x = 1/8.226 = 12.16% probability that p = 0.75 vs the alternative hypothesis of p=0.5 with a probability of 88.8% of the aunt not having the ability

if, instead the prior odds ratio was 1:1 (a priorly believed probability of 50%), after multiplying by the likelihood, you'd end up with a posterior odds ratio of 1/13.7 or 0.072993

now solving (1-x)/x = 0.072993, we arrive at x = 93.2% probability that p = 0.75 vs the alternative hypothesis that p = 0.5

note that 100%-93.2% = 6.7%

so in the first case a prior probability of 1% that the aunt has the ability is increased to 12.2% by the data likelihood

in the second case, the prior probability of 50% that the aunt has the ability is increased to 93.2%

$\endgroup$
  • $\begingroup$ the pdf author should be comparing 12.2% (case 1 ... up from a prior 1%) chance to 93.2% (case 2 ... up from a prior 50%) chance that the aunt has the ability $\endgroup$ – phdmba7of12 Sep 11 '18 at 21:59
  • 1
    $\begingroup$ thank you so much for the answer ! I am almost there. see edit, please walk me through the now solving (1-x)/x = 0.072993, we arrive at x = 93.2% & so (1-x)/x = 7.226 or x = 1/8.226 = 12.16% probability. need a hand hold step by step! LOL see my edits for the chart i am building to do this / where i am. TY $\endgroup$ – learnAsWeGo Sep 12 '18 at 5:17
  • $\begingroup$ if we called the odds ratio $r$, then $$(1-x)/x = r$$ solved for $x$ is step1. $$xr=1-x$$ step2. $$x(r+1)=1$$ step3. $$x=1/(r+1)$$ $\endgroup$ – phdmba7of12 Sep 12 '18 at 5:31
  • 1
    $\begingroup$ TY I will be throwing up questions as I work through this paper. you will probably be able to gain +500 points from this trail of questions LOL $\endgroup$ – learnAsWeGo Sep 12 '18 at 15:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.