Hello Guys

I have come across a research paper from a YouTube video I watched.

In the Abstract of the paper it says something that got me slightly confused, perhaps it is because of my limited English. Here is what it says:

Significantly greater hippocampal atrophy was observed for participants reporting a life-changing religious experience. Significantly greater hippocampal atrophy was also observed from baseline to final assessment among born-again Protestants, Catholics, and those with no religious affiliation, compared with Protestants not identifying as born-again.

Notice where I've made bold. Does that not make the research pointless ? The bit about people with no religious affiliations also having "greater hippocampal atrophy" !!

Or do they mean, they are religious, but they are not party of any organised religion ?

I would greatly appreciate some help in understanding this. Finally .. How can I tell if any research paper is peer-reviewed or not ?

Thank You.

Well, it has subjects self-identifying a subjective experience and then tries to correlate particular morphologies to those self-identified experiences. It's not gonna be too strong unless confounders associated with subject self-identification are strictly accounted for.

Hmm sorry piscator, can you some how put that into simpler English .. perhaps via examples ?

Sorry about this, I think I understand what you mean, but I want to be absolutely clear that I do.

I think this is data dredging. A reasonable sized study was set up to follow MRI changes associated with depression. A study of religious beliefs has been tacked on with no obvious objective and no clear hypothesis.

There were few catholics or individuals of no religious affiliation - so little statistical power.

The statistical analysis was performed for each hippocampal lobe independently and any positive result accepted.

P values for catholic, born again and non religious (<0.05) are not within the usual realms for a scatter gun approach like this where many analysis are performed hoping to find something publishable.

This is what I would call hypothesis generating - of no value without prospective validation.

Interesting. So you're saying it isn't too good, because the religious belief study wasn't the main aim of the research ?

Ciwan wrote:Interesting. So you're saying it isn't too good, because the religious belief study wasn't the main aim of the research ?

Sort of.

Statistics work provided correctly applied. All sorts of associations can be observed, but the implications can only be deduced if the study is constructed to deal with confounders. Thus A & B may be linked because C causes A and also B. Unless the study is constructed to address the precise question being investigated all sorts of things can appear causatively related.

Secondly the study was not powered to confirm most of the associations reported, in part because the study was not constructed to ask these specific questions.

Third any single studies findings could be chance alone - so without previous data showing that a relationship exists and a study powered to test whether such a relationship really exists, you simply have a possible association.

Cool thank you Pebble.

Temporary tangential diversion ...

An old, but still useful, book that will be educational, is William J. Reichmann's Use and Abuse of Statistics, ISBN 0140207074. This will provide some classic examples of failure to apply statistical inference rigorously, with worked examples.

Oh, and one major source of error involves using time as a linkage variable. Just because two sets of events happened over the same time period, doesn't mean that they were causally linked. If time is involved in any purported 'correlation', always be suspicious.