|
| Yea it measures fit. And since we don’t know the future, it’s the best predictor in my opinion of the future.
If the r2 was really low the model is worthless. A high r2 shows a really good fit model
What are we building these models for if not to assess future outcomes?
I’ve often thought about how in stat class it seemed too often to get hung up on the semantics of a thing.
Like “ correlation doesn’t prove causation”
And the analogy they always used was shark attacks and ice cream sales. They both trend up together and are highly correlated. But one doesn’t cause the other. What happens is summer comes around and more people are swimming at the beach and buying ice cream. However increased shark attacks doesnt strictly speaking mean an increase in ice cream sales.
But say you didn’t know what time of year it was. But you have data showing shark attacks increasing, you could make a reasonable assumption that ice cream sales will also increase.
Even tho one doesn’t cause the other, it doesn’t mean the correlation is useless information. | |
|
Old crop corn!
