5 ms·
A proposed solution, for the critique of any mathematicians here: Always split each channel into ten 10-percentile bands, instead of just taking an overall ave
by srgseg 15y ago
A proposed solution, for the critique of any mathematicians here:
Always split each channel into ten 10-percentile bands, instead of just taking an overall average. Then model total expected revenue per channel based on these bands.
Perhaps as few as three 33-percentile or five 20-percentile bands would be enough if the number of samples is low when first testing a channel.
- eru 15y agoPerhaps, if you are still fresh in integration and measure theory, you can go for a continuous solution, that doesn't require a choice of bands.
- _delirium 15y agoYou don't have access to the "real" continuous distribution, though, but only a finite sample of points from it. Most non-parametric ways of modeling that are going to require some choice of smoothing parameter, either something like a histogram bin width, or a kernel-regression bandwidth (in either case, you can use the data to choose one, using cross-validation).
- eru 15y agoHow about working with the distribution function instead of densities? That way you still have the sampling, but you don't have to decide on a bin size.