4 ms·
> offers clear-cut answers That is a huge exaggeration in the title of the press article. It's a very simple model y(t)= logit^(-1)(logit(y(0)+A*t+noise) and
by gus_massa 5y ago
> offers clear-cut answers
That is a huge exaggeration in the title of the press article. It's a very simple model
y(t)= logit^(-1)(logit(y(0)+A*t+noise)
and it somewhat matches the changes in the last 10 year, but the match is not as perfect as the title claim. The size of the change in many case is very different, and in some cases it's in the wrong direction. (Note that the authors of the papar neither claim a perfect match.)
They give 3 possibilities for A:
A=0, i.e. only noise, that is a very bad model
A=an arbitrary constant. I think this can be measured using previous surveys
A=a constant based in "argument advantage" that means "the sense that this opinion is easier to justify by generally acceptable kinds of arguments than the opposite opinion is."
The definition of "argument advantage" is very fuzzy, and "generally acceptable kinds of arguments" is an arbitrary criteria that depends on what the people think at the time and changes from time to time and form country to country.
It's not surprising that what people think is a acceptable argument and other moral opinions influence each other. So I think they are just trying to estimate the constant A in the model in a very indirect and imprecise way.
- PaulHoule 5y agoYeah, I read a book in the 1980s about how you can predict any social time series with logistic models. Maybe there is a better way to do it but I always fought with https://en.wikipedia.org/wiki/Levenberg%E2%80%93Marquardt_algorithm https://en.wikipedia.org/wiki/Levenberg%E2%80%93Marquardt_al... and got a fit that looked pretty good but when the expected "event" had happened it had nothing to do with reality. When the LEI was required to trade certain derivatives in the US the registration rate for LEIs formed a nice logistic curve but when they announced "no LEI, no trade" for more common things in the US my post-hoc curve fit was totally wrong because 70% of the registrants registered at the last minute.