Why no international statistical testing?

#1
I am wondering why it's generally not done to conduct a statistical testing comparing prevalence rate between two different countries even they have used the same surveys/methodology?

We can adjust for potential confounders such as age to account for the difference in age distribution between the two countries. I've got some ideas in my mind but not very convincing, so would like to hear from you all.

Thanks in advance.
 

CB

Super Moderator
#2
It can be and is done, but the issues are more complex than accounting for confounds in terms of demographic differences between the samples - depending on the study field, one also often needs to ensure that the measuring instrument is invariant between the two samples (e.g. differences in extreme and acquiscent response styles). Invariance testing means testing that the measuring instrument's items are related to the trait being measured in the same way and to the same degree across the samples (usually using structural equation modeling).

Good discussion of some of these issues here:
Cheung and Rensvold, 2000

What field are you, in by the way?
 
#3
Thanks CowboyBear,

So are you saying that the residents residing in different countries may interpret and respond to the exact same question inherently differently? That's one thing I have in mind but wouldn't a validated questionnaire should have taken care of that bias?

The problem is if there is so many underlying hindrances for the international comparison, why do we compare the prevalence rates qualitatively to begin?

I am in Epidemiology, doing my thesis on international prevalence of a respiratory disease among adults. I am getting my defense done soon, so trying to come up with possible questions to be asked.

I will look over the paper you refer to, thanks!!! :tup:
 

CB

Super Moderator
#4
So are you saying that the residents residing in different countries may interpret and respond to the exact same question inherently differently? That's one thing I have in mind but wouldn't a validated questionnaire should have taken care of that bias?
Exactly. A validated questionnaire helps with this only if the measure has been validated in terms of its invariance across the cultures being studied - which requires quite sophisticated methodological analyses.

I am in Epidemiology, doing my thesis on international prevalence of a respiratory disease among adults. I am getting my defense done soon, so trying to come up with possible questions to be asked.
Ah, I'm in the social sciences, there are some differences here obviously. E.g. if you can measure prevalence via an established physical diagnostic procedure (e.g. lab tests or whatnot) you wouldn't really be worrying about measurement invariance as much - though if prevalence is being estimated by self-report questionnaire, the issues would crop up again.

There do definitely seem to be quite a number of international epidemiological comparisons in medicine - a Google Scholar search for 'international comparison of prevalence' brings up 1.7 million hits, with a lot of medical studies in the first couple of pages anyway!