Thank you Mike!

Thanks Dave!

It is also (more likely) their browser, switching to an alternative usually does the trick.

Just looking at the backend now. Appears to be a user spike issue. I’d suggest getting them to pair up for the time being. This has been reported before, we are waiting on some funding to increase capacity.

That’s amazing to hear! Do let me know if your class has any comments or run into any issues.

... (though something similar could exist and I am unaware?). You are of course right, doing something like this is ultimately a trade-off between measurement error and interpretability.

It certainly seems better than studies that stick SA clusters into a MLM and calls it a day.

Promising in the sense that this new method appears to provide the ability to make such a tradeoff. The bootstrapping provides a relative threshold mechanic to assess the relative cluster stability. I think this is what has been missing from models that use SA clusters...

There are attendance monitoring systems in place. Though I have no idea why any member of academic staff would use them. I am not a immigration enforcement officer and realistically that is why attendance is recorded.

Thank you for the suggestion. Off the top of my head the differences in R^2 are very small, but I will re-run some analysis to test for sure.

My general conclusion also - though I do feel there is some interesting effect size differences going on when duplicating the 2007 paper's model on auth-lib scales?

If all are measuring the same thing, I suppose that opens the "Is it measuring status or social distance?" conversation once more.