1
$\begingroup$

For some context: I've been looking at some meta-analyses on factors that explain happiness (well-being), because I wanted to check the happiness-claim of a certain philosophy and want to be happy in general (who doesn't :). Now of course these meta-analysis reveal different covariables for happiness with different correlation coefficients.

So I wonder: Is the size of the correlation coefficient the only thing I need to look out for in order to understand what creates the most happiness long term, according to these studies? Or do I need to look at the effect size? Or what else to look out for? I guess I want to understand what I should do to get the most amount of long term happiness out of one hour spent of my time?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Correlation coefficients are types of effect sizes, so correlation isn't really distinct from effect size. Look also for the significance of the coefficient (with a p-value or some other statistic). Also look out for confounding variables--correlation isn't causation, so insight from multiple regression models would be useful.

Improve this answer
$\endgroup$
4
  • $\begingroup$ thanks for your insights! I guess the question is: Is correlation and effect size proportional? Meaning If you have higher correlation, you always have higher effect size? Or are the examples or situations where correlation might be lower, but effect size might be BIG? For example: Lets take sleep hours as input variable and feeling of fatigue as output. Everyone who has no sleep would have high fatigue feeling - low variance in that section. But a lot of people have enough sleep and still feel fatigue. This might be an example for something with a high effect size but lower corrrelation? $\endgroup$
    – CrossVal_Asker
    Commented Jul 9 at 17:14
  • $\begingroup$ Effect size is a general category of measurements that indicates things about two variables. Correlation coefficient is one specific type of effect size. So you have to define effect size first. Is it the correlation coefficient? Is it the means difference? Is it the odds ratio? If it's the correlation coefficient, then the coefficient IS the effect size. If it's something else, then it may measure a different variable relationship, so it may be different from the correlation coefficient. $\endgroup$
    – Steven Morrison
    Commented Jul 9 at 17:24
  • $\begingroup$ Good question about effect size. I realize I’m not entirely sure what effect size means. What I'm really asking is for a practical application: If I do one more unit (or even better one extra hour) of this, how much change will it cause in the output variable? I mean that would be the practical knowledge here, right? $\endgroup$
    – CrossVal_Asker
    Commented Jul 9 at 17:29
  • $\begingroup$ If you're asking whether you'll get an increase in Y if you do more of X, and your hypothesis is that you will because X and Y are highly positively correlated, the fault in your logic is that correlation is not causation. For example, assume that sunscreen use of positively correlated with skin cancer--as more people use sunscreen in a population, the rate of skin cancer increases. Does sunscreen CAUSE skin cancer? No. The confounding variable is sun exposure: increased rates of sun exposure result in both more sunscreen use and more skin cancer. $\endgroup$
    – Steven Morrison
    Commented Jul 9 at 20:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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