The DHS Program User Forum
Discussions regarding The DHS Program data and results
Home » Topics » Child Health » Odds ratios vs Marginal effects
Re: Odds ratios vs Marginal effects [message #9382 is a reply to message #8982] Tue, 22 March 2016 20:29 Go to previous messageGo to previous message
user-rhs is currently offline  user-rhs
Messages: 132
Registered: December 2013
Senior Member
Reduced-For(u)m wrote on Fri, 22 January 2016 13:04

Dear Economist,

Please continue to estimate marginal effects. They are so much easier to interpret.

Another Economist

*Ok, for real: it is just a matter of disciplinary convention, and any number of methods can work for any number of problems. You should feel free to estimate whatever kind of effects make sense in your research (and in your discipline).

As someone who dabbles both in epidemiology and economics, I respectfully disagree with R-F. They are *both* equally easy to interpret. I have, on occasion, chosen to present odds ratios instead of marginal effects when the marginal effect was abysmal (~0.00005) and the odds ratio was much more impressive ("1.5 times more likely"). I suspect others do the same. Sometimes the funders don't want to see the "true results," and you have to be...clever in packaging your results. It's not exactly "dishonest" so long as you're not making up data or numbers. I've seen this done even in highly cited studies on interventions that claim to have large protective effects against disease X, but the biological plausability of the intervention is suspect, and potential moral hazard brought about by the intervention puts the "clinical significance" of the intervention under even greater scrutiny. To be fair, I had never heard of marginal effects until I started working under an economist, who, not incidentally, loathes odds ratios.


NB: In all seriousness--yes, I agree with R-F's sentiments above. How you present your results depend on the research, your field/discipline, and your target audience. Public health and policy people might prefer odds ratios because "two times more/less likely" is easier to digest than actual percentage point increases/decreases

[Updated on: Tue, 22 March 2016 20:57]

Report message to a moderator

Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Read Message
Previous Topic: Region Variables
Next Topic: Assessing inter-generational effect of malnutrition
Goto Forum:

Current Time: Mon May 17 09:11:06 Coordinated Universal Time 2021