Meta-analysis using different units of measurement

[Emily Blan]

Hello, I'm compiling data for a meta-analysis regarding enzyme concentrations in a patient group compared to a control group that are quantified in papers using different units of measurement, for example μmol/l, nmol/ml, pg/ml, ng/ml etc. My question seems rather elementary, but I can't find sufficient evidence elsewhere to back up my thinking. My inclination is to forego converting the values to a standard unit, but rather to use the standardized mean difference as a 'neutral' measure. Is this correct?

[Marcos Almeida]

Welcome to the Stata Forum / Statalist,

I gather your decision to rely on the SMD is quite reasonable.

[Emily Blan]

Thank you Marcos!

[Bruce Weaver]

Hi Emily. The formula for the standardized mean difference will have some kind of pooled SD as the denominator, I suppose. This means that two studies showing the same mean difference could show very different standardized mean differences if their pooled SDs differed. Or two studies with very different mean differences could show exactly the same SMD given the right combination of SDs. This makes me uncomfortable with standardized measures. My initial inclination would be to convert the raw mean differences to a common scale, if possible. YMMV.

[Bruce Weaver]

Here's an article that discusses some of the problems associated with standardized measures of effect size:

• Baguley, T. (2009), Standardized or simple effect size: What should be reported? British Journal of Psychology, 100: 603–617.

HTH.

[Tim Morris]

Hi Emily
Rebecca Walwyn and Chris Roberts' recent article is about meta-analysis of standardised mean differences and may be helpful.
http://onlinelibrary.wiley.com/doi/1....7186/abstract
Tim

[Emily Blan]

I'm sorry I'm so late to reply but thank you so much everyone!