3
$\begingroup$

I'm peer-reviewing a manuscript for a psychology journal in which I believe the authors have mixed up odds-ratio and risk-ratio. They are being so stubborn in their insistence that they have not mixed them up that I feel like I need a sanity check. I have written a paragraph intended to encapsulate the whole issue. Could anyone please simply endorse the following paragraph or explain how it needs improving? I have edited some words (e.g. factor name) to preserve anonymity of the manuscript.

Quoting directly from the manuscript: "Inspection of the regression model revealed that the mean population odds ratio is estimated to be 1.90 (95% CI: 1.00, 3.58), suggesting that [participants] are 1.90 times more likely to [do X in] the [A condition] than the [B condition]." This is from a straightforward binary logistic regression with a categorical factor [A vs. B] which can be [A] or [B]. Let's look at a standard definition of Risk Ratio (https://en.wikipedia.org/wiki/Relative_risk): "relative risk or risk ratio (RR) is the ratio of the probability of an event occurring (...) in an exposed group to the probability of the event occurring in a comparison, non-exposed group". In this context, "times more likely" (manuscript) is the same as "ratio of the probability of the event occurring in a ... group to the probability of the event occurring in a comparison ... group" (wikipedia's definition of RR). The first worked example of RR in wikipedia in fact uses the exact same phrase "times more likely". This clearly demonstrates that the authors are referring to an OR as if it was an RR. The internet is full of easily found accurate pages explaining that OR is not the same thing as RR (but that they are easily confused). QED.

$\endgroup$
3
$\begingroup$

You are of course right and it is a common mistake to describe an odds ratio like a relative risk ratio. I would suggest that it would be helpful to propose a more appropriate phrasing to them such as "suggesting that the odds of [participants] [doing X in] [A condition] is 1.90 times higher than in [B condition]." Once the authors realise that is all they would have to do, hopefully this is not too much of a discussion.

$\endgroup$
  • $\begingroup$ Thanks very much Björn. Sanity check very helpful. For the record (as I will direct the editor's attention to this page) I note that from your profile you are a "Biostatistician working in pharmaceutical drug development" and can therefore be considered a probably reliable source. $\endgroup$ – Amorphia May 18 '16 at 11:05
  • 1
    $\begingroup$ An internet forum is not a reliable source. Also, the fact that a brilliant person says something does not proof that that statement is correct. Instead I would just stick to proposing the alternative, and leave the decision to the editor. In the end, it is her or his decision, not yours. $\endgroup$ – Maarten Buis May 18 '16 at 11:46
  • $\begingroup$ Fair point - just looking for a little corroboration here. $\endgroup$ – Amorphia May 18 '16 at 15:09
  • $\begingroup$ Hi. I know i'm a little late to this question, but just to be clear. It is not correct to use the terms 'more likely', 'less likely', etc when describing odds or odds ratios? 'Likely' can only be used to describe probabilities? $\endgroup$ – RNB Oct 12 '16 at 8:15
  • $\begingroup$ Most people understand "1.5 times more likely" to mean "a risk ratio of 1.5", so in a sense, yes, I feel that "likely" is most commonly used to describe probabilities. Using it for anything else feels potentially very misleading to me, especially because odds ratios are typically larger and this unclarity might be seen the exaggerate an effect. "More likely" and "less likely" without specific numbers are of couse fine to describe "higher odds" and "lower odds", because probabilty to odds is a one-to-one transformation. $\endgroup$ – Björn Oct 12 '16 at 12:42

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.