7 ms·

These guys' models can have multiple solutions. To me, that would indicate a problem with the model. But they are happy enough just to select the most optimisti

by alimw 5y ago

These guys' models can have multiple solutions. To me, that would indicate a problem with the model. But they are happy enough just to select the most optimistic solution.

teorema 5y ago
I have to read the paper thoroughly but I've come to the opinion that multiple solutions aren't in of themselves a problem if they are incorporated into inferences. A lot of Bayesian inference can be thought of as involving solution spaces, for instance, where you're assigning a probability to each solution. Not disagreeing with you, just wondering out loud if it could be revisited with other methods.
epistasis 5y ago
I don't do financial modeling, but in all other areas of modeling it's definitely the case that there are many different "solutions" by which I'm guessing you mean local optima. Having multiple solutions is somewhat equivalent to having different states of matter in physics. And economies definitely have very different modes, or states, that we call recessions or booms. Being able to have these sorts of local optima would seem to be necessary foe any model of an economy. But I may not be understanding you, because we are using the same words in different ways...
- carlob 5y ago
  I took a class of econophysics in grad school and one of the models with studied was a very simple generational model to explain inflation. It was originally discarded because it had an unrealistic solution with very high inflation, however it turned out to be a very good model for how hyperinflation can arise with a relatively small change in parameters. This was made particularly poignant given that the professor was Argentinian.
- alimw 5y ago
  This turns out to be a review paper by the same authors which only mentions the model I wanted to criticise. It's unlikely that one of the original papers will ever turn up on HN though! I was surprised to see this. As I recall they began by adopting a structural model for debt valuation that assumes no dependence on any external assets other than the debtor's. Then they reintroduce full n-dimensional dependence through a different mechanism they made up to try and force the 1-dimensional solutions into a state of consistency. It's a bit of a mess. Actually you're right. The presence of multiple solutions is not in itself enough to make me hate the model. The analogy relating different economic regimes to different phases of matter is potentially a good one.
XnoiVeX 5y ago
As statistician George Box said, "All models are wrong, but some are useful"

by alimw 5y ago

These guys' models can have multiple solutions. To me, that would indicate a problem with the model. But they are happy enough just to select the most optimistic solution.

teorema 5y ago

I have to read the paper thoroughly but I've come to the opinion that multiple solutions aren't in of themselves a problem if they are incorporated into inferences. A lot of Bayesian inference can be thought of as involving solution spaces, for instance, where you're assigning a probability to each solution. Not disagreeing with you, just wondering out loud if it could be revisited with other methods.

epistasis 5y ago

I don't do financial modeling, but in all other areas of modeling it's definitely the case that there are many different "solutions" by which I'm guessing you mean local optima. Having multiple solutions is somewhat equivalent to having different states of matter in physics. And economies definitely have very different modes, or states, that we call recessions or booms. Being able to have these sorts of local optima would seem to be necessary foe any model of an economy. But I may not be understanding you, because we are using the same words in different ways...

carlob 5y ago
I took a class of econophysics in grad school and one of the models with studied was a very simple generational model to explain inflation. It was originally discarded because it had an unrealistic solution with very high inflation, however it turned out to be a very good model for how hyperinflation can arise with a relatively small change in parameters. This was made particularly poignant given that the professor was Argentinian.
alimw 5y ago
This turns out to be a review paper by the same authors which only mentions the model I wanted to criticise. It's unlikely that one of the original papers will ever turn up on HN though! I was surprised to see this. As I recall they began by adopting a structural model for debt valuation that assumes no dependence on any external assets other than the debtor's. Then they reintroduce full n-dimensional dependence through a different mechanism they made up to try and force the 1-dimensional solutions into a state of consistency. It's a bit of a mess. Actually you're right. The presence of multiple solutions is not in itself enough to make me hate the model. The analogy relating different economic regimes to different phases of matter is potentially a good one.

XnoiVeX 5y ago

As statistician George Box said, "All models are wrong, but some are useful"