jerome wrote:Yep agreed VK. This is what I don't get. I can't see why their hypothesis arose, and why they suggest a logistic curve. If anything it rather suggests backwards hypothesis formation: however I'm pretty certain they had a reason, I just don't understand the logic behind the maths in any way.

j x

I don't have access to their paper so I am going by what you quote on your blog.

The logistic curve follows straightforward from the assumptions given. That's textbook stuff.

I can't say whether the assumptions are in any way justifiable but I have read on social psychology in the wake of the Bem debacle. "Backwards hypothesis formation" is normal practice in that field. If it looks like that, then that's probably the case.

I don't think you can do a significance test on that data. For that you'd have to calculate a probability.

The curve does not tell you anything about probabilities. When you fit the curve you can calculate a standard error from the residuals which you could use for that purpose. But I don't think you are justified in assuming the same SE for all diaries.

There are a couple ways to measure goodness-of-fit or to validate curves in other ways but no real tests as such. Different data sets have different errors and different confounders.

But I don't really know much about time series analysis. You'd probably want an econometrician for an authoritative answer on what you can do in such a situation.