Hello Gijs,

ES (effect size) may be quantified in a number of different metrics. The most suitable depends on your specific research aims.

In multiple linear regression, you could use:

1. The model R-squared (R^2) as an ES

2. Cohen's f-squared as an ES (= R^2 / (1 - R^2); a signal to noise ratio )

3. For an individual IV (independent variable), the estimated standardized regression coefficient (sometimes called "beta") could be used as an ES

4. The difference in R-squared between a model that included a specific IV and that for a model that excludes the specific IV could be used as an ES (this is akin to hierchical regression models).

I'm sure that others could add to this list, but perhaps that's enough to get you started.

Good luck with your work.

Hi Gijs,

It also depends on what the other studies in your MA report as effect size. Are you loooking for correlation coefficients for a specific IV-DV relationship? All of what David says is true, but not so common to use as effect sizes (in management).

In that case, it is more or less common practice (in management) to compute partial correlations from regression models in case zero-order (normal) correlations are not reported. For a partial correlation, we derive the correlation between x and y (controlled for a vector of z other variables in the model), from the regression model directly. When you look at the formula form a partial correlation, its easy to compute usually, since t = ES/SE, most of this information should be given or can be derived.

Kind regards,

Robert Suurmond

Gijs Tempelman This is a very interesting and useful question in scientific research. Simply to conduct the effect size of the study, the correlation coefficient for the study variable must be obtained, and by using Cohen's test the effect size can be obtained. The attached table displays equivalent effect sizes expressed as Cohen's d values and Pearson correlation coefficient.