[hahnjo]

No description provided.

[hahnjo]

Here I wasn't sure what to print. Initially I wanted to do "errors", but there are several definitions so I kept it to the minimum... Let me know if you would like to see something else printed.

[hageboeck]

I would suggest mean(x) and variance (or stdDev), which should be close to 1.

[github-actions]

Test Results

    23 files      23 suites   3d 21h 21m 44s ⏱️
 3 862 tests  3 859 ✅ 0 💤 3 ❌
79 609 runs  79 605 ✅ 1 💤 3 ❌

For more details on these failures, see this check.

Results for commit 2b287fe.

[hahnjo]

ping

[hageboeck]

I would suggest to extend RProfileBin, so that is can compute mean, variance, stdDev and standard error of mean.
For the plot, I suggest to show the means instead of the bin content to underline the "dimensionality reduction" that's the main job of the profile.

[hageboeck]

Suggested change

	ROOT::Experimental::RProfile profile(40, {0.0, 20.0});
	ROOT::Experimental::RProfile profile(axis);

[hageboeck]

It's more interesting to vary the distribution along y depending on x:

Suggested change

	// Create a first normal distribution with mean 10.0 and stddev 4.0.
	std::normal_distribution normal1(10.0, 4.0);
	// Create a second normal distribution for the value.
	std::normal_distribution normal2(40.0, 10.0);
	for (std::size_t i = 0; i < 1000; i++) {
	double x = normal1(gen);
	double v = normal2(gen);
	profile.Fill(x, v);
	}
	// Fill uniformly along x.
	std::uniform_distribution normal1(0.0, 21.0);
	for (std::size_t i = 0; i < 1000; i++) {
	double x = normal1(gen);
	// In y, use a normal distribution with a mean that depends on x.
	std::normal_distribution normal2(x, 10.0);
	double y = normal2(gen);
	profile.Fill(x, y);
	}

[hageboeck]

Adding Variance, StdDev and the error of the mean would be interesting, too.

[hageboeck]

I suggest to plot the mean for each bin in x. The scale might need a bit of adjustment, but we should see a rising line.

Suggested change

	const auto &content = profile.GetBinContent(bin);
	const auto &content = profile.GetBinContent(bin).ComputeMean();

[hageboeck]

I found this to work reasonably well:

   for (int row = 8; row >= 0; row--) {                                                                                 
      auto print = [&](ROOT::Experimental::RBinIndex bin) {                                                             
         const auto &content = profile.GetBinContent(bin);                                                              
         const int targetRow = content.ComputeMean() / 25. * 8;                                                         
         std::cout << ((row == targetRow) ? '-' : ' ');                                                                 
      };     

Output:

entries = 1000
binned mean = 12.3241, stddev = 7.18306
 |                                        | 
 |                                        |-
 |                                     ---| 
 |                                -----   | 
 |                         -------        | 
 |                  -------               | 
 |            - ----                      | 
 |      ------ -                          | 
 |------                                  | 

value mean = 12.3319, stddev = 7.33413
underflow bin: entries = 0, mean = nan
bin #5: entries = 20, mean = 2.56564
bin #10: entries = 22, mean = 5.05397

[hageboeck]

Suggested change

	/// Profile histograms with RProfile.
	/// Profile histograms with RProfile.
	///
	/// Profiles can be used for dimensionality reduction by representing an N-D distribution in N-1 dimensions using summary statistics such as mean and variance.
	/// In this tutorial, a 2-D distribution uniform in x and normal in y with changing mean is represented by its mean values in each bin along x.

[hageboeck]

I would suggest mean(x) and variance (or stdDev), which should be close to 1.