4 ms·
> You can decide for yourself whether an error like this—7 minutes out of 9 hours—makes the real data “close” to a sine wave. I think the most interesting part
by jmward01 2y ago
> You can decide for yourself whether an error like this—7 minutes out of 9 hours—makes the real data “close” to a sine wave.
I think the most interesting part of this article is the bit at the end. What really is 'close'? We have so many 'rules of thumb' here but a real definition to target is elusive. Do you go off of pure utility? 'using this definition achieves my requirement to be corret xxx% of the time and now I make money using it' Or do you go off of something more like information content: 'when plotting the error it conforms to a normal curve...' Anyone got a good rule for 'close'?
- abnry 2y agoI don't think there will ever be an answer to this. There's a similar problem with choosing a threshold for a decision. The common thing to do is to make a RoC curve to compare the trade-off between true positive rate and false positive rate. If you have a system you can reason about completely, then sometimes you have a number that gives the absolute answer. Say you get error below floating point resolution. But I guess it's otherwise what is perceptible or meaningful, either in quantity or percentage. A penny is not a life changing amount of money and something happening 0.01% of the time is rare enough to be tolerable.