The black dots (and dashed lines) represent the observations and the bars are the predictions from the model. The predictions "hang" from the observations. A perfect model will be one where the hanging bars touch the dashed grey lines at zero. Instead of the frequency, we plot its square root. This makes it easier to compare observed and expected frequencies
even for low frequencies.
The previous figure is from an example in BAP3. We should create a more Bayesian version and represent the uncertainty for the bars.
Rootograms are a visual model checking tool for count models. You can read more here and https://fromthebottomoftheheap.net/2016/06/07/rootograms/ here https://arxiv.org/abs/1605.01311
The black dots (and dashed lines) represent the observations and the bars are the predictions from the model. The predictions "hang" from the observations. A perfect model will be one where the hanging bars touch the dashed grey lines at zero. Instead of the frequency, we plot its square root. This makes it easier to compare observed and expected frequencies even for low frequencies.
The previous figure is from an example in BAP3. We should create a more Bayesian version and represent the uncertainty for the bars.