0
$\begingroup$

Let's say we found strong correlation between meat consumption and life expectancy. We also know there's correlation between GNI and meet consumption and between GNI and life expectancy. Is there a way to mitigate influence of GNI in our first correlation so connection between meat consumption and life expectancy become more clear?

*GNI stands for Gross National Income

$\endgroup$
  • 2
    $\begingroup$ I think you are probably looking for either partial correlation or some form of multiple regression although without more detail I am not clear which you want. $\endgroup$ – mdewey Nov 22 '18 at 15:42
  • $\begingroup$ @mdewey what details should I provide, I'm just a beginner in stats, sorry $\endgroup$ – imbolc Nov 22 '18 at 15:45
  • 1
    $\begingroup$ Specifying the scientific question might help. Are you interested in modelling one of the variables as a function of the others or just looking at their relationship? It might also help to know what GNI is. $\endgroup$ – mdewey Nov 22 '18 at 15:53
  • $\begingroup$ @mdewey ah, sorry, its Gross national income en.wikipedia.org/wiki/Gross_national_income I think I'm looking at their relationship, trying to understand how meat consumption itself related to health. But if meat consumption strongly correlates with people wealth, which probably correlates with healthcare and other variables which can be the real cause of extended life span. So I'm wondering if it's possible somehow mathematically "substract" their influence. $\endgroup$ – imbolc Nov 22 '18 at 16:36
  • 1
    $\begingroup$ Multiple regression would be worth investigating then. $\endgroup$ – mdewey Nov 22 '18 at 18:15
1
$\begingroup$

Just so this does not join the pile of unanswered questions.

You can use multiple regression here. You would model life expectancy (LE) as a function of meat consumption (MC). So LE is on the left hand side, MC on the right hand side. You would then include GNI. The coefficient of MC now would give you the effect of MC on LE allowing for GNI sometimes expressed as holding GNI constant.

Some caveats.

This is an ecological analysis so what it is saying is that countries with high meat consumption have higher or lower life expectancy. It does not tell you what happens for individuals. Similarly it is the effect of living in a country with high or low wealth not having that personally. In a country with high inequality most people might be very poor even though the GNI is high.

The direction of causality, if any, is not clear here either. It seems reasonable that what you eat affects when you die but it could be that people who for other reasons live for a long time happen to really like eating meat.

$\endgroup$
  • $\begingroup$ > It does not tell you what happens for individuals. @mdewey I don't quite get it, it's obvious about wealth as the gap can be huge. But in case of lifespan, don't you think deviation is pretty modest, no one lived even for 150 years as I know. So why do you think we can't transfer the relationsheep into individual level here? The only factor coming to my mind is child mortality, but we can easily exclude countries-outliners. $\endgroup$ – imbolc Nov 25 '18 at 12:52

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.