docs/04-histdist.md

In ppernot/ErrView: Analysis of error sets

HistDist module

This module enables to plot the distribution of errors (as histogram) and to explore the trends of error sets.

Controls

Method

• Select: select the error set

• Remove glob. outl.: remove global outliers (if any) defined in module Outliers

Histogram

• # Hist classes: number of classes in the histogram (defaults to Scott's method when slider set to 0)

• Normal fit: draw a normal fit to the errors histogram

Trend line

• Plot trend line: plot the trend line for a polynomial defined by Trend degree

• Correct trend: correct the errors from the trend line

• Trend degree: degree of the polynomial trend line

Errors vs. Data

• Bland-Altman: plot Bland-Altman statistics (2.5% and 97.5% interval + mean) with their bootstrapped uncertainty

• Tag outliers: plot the tags of points out of the 95% confidence interval defined by the Bland-Altman lines

• Glob. errror scale: scale the y-axis from the full dataset (all methods) or from the current one

Plots

The left panel presents a histogram of the errors and oprionally a normal fit (same mean and variance). The histogram reflects the transformations to the errors set (outliers removal, trend correction...)

The right panel plots the errors as a function of the calculated data in order to visualize trends that might be corrected to improve predictions.

ppernot/ErrView documentation built on Jan. 30, 2022, 6:59 a.m.