Not sure what “attenuation” means in this context.
It’s probably correction for attenuation: ‘Correction for attenuation is a statistical procedure … to “rid a correlation coefficient from the weakening effect of measurement error”.’
Ah, thanks! So as a practical matter it seems like we probably shouldn’t correct for attenuation in this context and lean towards the correlation coefficient being more like 0.26? Honestly that seems a bit implausibly low. Not sure how much stock to put in this paper even if it is a meta-analysis. Maybe better to read it before taking it too seriously.
I’d correct for attenuation, as we care more about getting the people who in fact will perform the best, rather than those who will seem like they are performing the best by our imperfect measurement.
Also selection procedures can gather other information (e.g. academic history, etc.) which should give incremental validity over work samples. I’d guess this should boost correlation, but there are countervailing factors (e.g., range restriction).
Oh interesting, I was thinking it would be bad to correct for measurement error in the work sample (since measurement error is a practical concern when it comes to how predictive it is.) But I guess you’re right that it would be reasonable to correct for measurement error in the measure of employee performance.
It’s probably correction for attenuation: ‘Correction for attenuation is a statistical procedure … to “rid a correlation coefficient from the weakening effect of measurement error”.’
Ah, thanks! So as a practical matter it seems like we probably shouldn’t correct for attenuation in this context and lean towards the correlation coefficient being more like 0.26? Honestly that seems a bit implausibly low. Not sure how much stock to put in this paper even if it is a meta-analysis. Maybe better to read it before taking it too seriously.
I’d correct for attenuation, as we care more about getting the people who in fact will perform the best, rather than those who will seem like they are performing the best by our imperfect measurement.
Also selection procedures can gather other information (e.g. academic history, etc.) which should give incremental validity over work samples. I’d guess this should boost correlation, but there are countervailing factors (e.g., range restriction).
Oh interesting, I was thinking it would be bad to correct for measurement error in the work sample (since measurement error is a practical concern when it comes to how predictive it is.) But I guess you’re right that it would be reasonable to correct for measurement error in the measure of employee performance.