Hans Dembinski

We seem to converge on `poisson_variance()`. I like that. It reads well and still conveys the idea that the variance could be computed in a different way. By the way,...

> I like that package because it teaches users how to handle their data properly, unlike ROOT where you can divide and subtract histograms and then still store them as...

degski on boost mailing list: > From what you are saying, and I have no knowledge at all in this matter [just reading what you say], it seems that a...

@degski Strictly speaking, there are infinitely many ways to compute variance, because it depends on the details of the process that generated the data for the histogram. Strictly speaking, we...

The Poisson variance has the advantage that it can be computed individually for each point. The multinomial variance can only be computed by looking at all bins, since you must...

> since variance cannot generally be considered a function of a single bin, you might prefer to make it a function of the histogram instead: h.poisson_variance(i), this would leave room...

Yes, but as a functor. ``` auto view = multinomial_variance_view(h); auto stat = view(i, j, k); ```

Hi, you were right to ask, there is something about this in the docs, but quite hidden and I didn't expect you to find it: https://www.boost.org/doc/libs/develop/libs/histogram/doc/html/histogram/rationale.html#histogram.rationale.comparison_to_boost_accumulators Boost.Histogram is designed for...

Bayesian Blocks is the most promising algorithm for automatic binning known to me: https://www.astroml.org/examples/algorithms/plot_bayesian_blocks.html It suffers from less issues than the other rules, but is not perfect either. It is...

That's cool, I should add this to the docs.