[doorhammer]

I think the data is a fascinating point for lively conversation and conjecture as long as you understand that you can only draw so much so much in the way of specific causation from it.

I didn't look at the raw data, but it also seems like some of the info might be inherently linked. Eloping might imply you don't regularly attend church, or if you do that you're marrying outside the church consent. Regularly attending church can bring with it a relatively stable set of defaults to invite to a wedding. A large attendee list also might imply affluence, or having taken longer to plan the wedding, implying a longer pre-marriage relationship. All just flippant conjecture, of course, but interesting.

I wonder if there's anything in the data you could use as a sort of halfway proxy for whatever you'd consider a 'happy' or contented relationship.

I grew up in a church environment as well, and I saw my share of marriages held together by social pressure, but I also wonder if some of the other marriages are better because a lot of people I knew very much had the idea that "this is forever, so I don't even have to think about it anymore." It wasn't even a point of consideration anymore. I wonder if that can relieve some level of psychological stress in a marriage that's not absolute Perfect, but pretty decent, instead of wondering if you missed some distorted version of "the one". (again more conjecture)

[Periodic]

It is possible to do these analyses and isolate a single variable regardless of the correlations with other variables. I'm hoping they did this but their methods section does not seem to indicate that they did.

They have a very large variable space (one variable per question) and they want to reduce that to the ones that are actually interesting. Hopefully they did some sort of dimensionality reduction to find the important variables, and I expect a lot of those new dimensions were comprised of many variables, for example, wealth and money spent on wedding are probably correlated.