Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Types of changes:

  • Added for new features.
  • Changed for changes in existing functionality.
  • Deprecated for soon-to-be removed features.
  • Removed for now removed features.
  • Fixed for any bug fixes.
  • Security in case of vulnerabilities.

[1.10.1] - 2026-03-03

Added

  • The PairwiseMatrixCharts class to precast module to visualize pairwise relationships between features in a dataset.
  • The Box plotter class to plotlib module for classical boxplot visualization using matplotlib native boxplot.
  • The penguins dataset to datasets module.
  • Images for each precast example.
  • Visual tests for Box plotter.

Changed

  • Default boxplot rcparams within daspi-dark and daspi mplstyle.

Fixed

  • Generating the list of subgroup values at SpreadOpacity class failed if some quantiles were the same, which could happen when there are many duplicate values in the data. This is now fixed by using all values of the quantiles as subgroup values instead of only the unique values.

[1.9.3] - 2026-02-10

Added

  • Static method by_resolution to FractionalFactorialDesignBuilder class to create instances by providing resolution.
  • Comprehensive tests for the new by_resolution method.
  • Step-by-step process drawings for Gage R&R analysis.
  • Enhanced documentation with additional information for 3s methodology.
  • Improved gage analysis documentation and guides.

Changed

  • GageRnRModel.uncertainties() method now returns concatenated uncertainties for better usability.

Fixed

  • Bug in axis parameter of groupby method of pandas that is deprecated in pandas v3.0.0, which caused errors in LinearModel class.
  • Bug in strings.py where getting USERNAME failed when using DaSPi on servers (e.g., Google Colab).
  • DOE import error, could not import get_default_generators

[1.9.0] - 2026-01-27

Added

  • The agreement parameter to RandomProcessValue class to define the sigma agreement for process capability calculations.
  • The 3s methodology guide to the documentation.
  • The gage analysis guide to the documentation.

Changed

  • The option name disturbances in LinearModel to covariates for better understanding.

[1.8.0] - 2025-11-04

Added

  • The DOE module with classes and functions for design of experiments.
  • The Factor class to DOE module.
  • The BaseDesignBuilder class to DOE module.
  • The FullFactorialDesignBuilder class to DOE module.
  • The FullFactorial2kDesignBuilder class to DOE module.
  • The FractionalFactorialDesignBuilder class to DOE module.
  • Comprehensive test coverage across the entire package, now at approximately 95%.
  • The property errors to ProcessEstimator class to show errors as percentage values.
  • 'About DaSPi' section to documentation.
  • Assert tests to check whether data are empty for SingleChart class.
  • The method check_current_data() to SingleChart class to check if current data is empty, throws a UserWarning if so.
  • The option to set a column name when generating descriptive statistics using describe() method of Estimator classes.
  • The option to set a column name when generating process descriptions using process_descriptions() method of Process classes.
  • The option to rotate, align and format tick labels. This option is applied to SingleChart, MultivariateChart and the underlying LabelFacets classes.

Changed

  • The property processes of ProcessEstimator class to a method.

Fixed

  • Creating uncertainties DataFrame for GageRnRCharts failed in certain cases.
  • The functions stdev_ci and variance_ci had incorrect calculations.
  • The stacked bars in Bar plotter were not always displayed correctly. With this fix, t_base is now calculated correctly.

[1.7.0] - 2025-07-01

Changed

  • The name of Estimator class to LocationDispersionEstimator and this class now also inherits from BaseEstimator class.
  • The property name descriptive_statistic_attrs of Estimator classes to attrs_describe.
  • The GageRnRModel now calculates uncertainties according to the VDA 5 standard.

Added

  • The BaseEstimator class to estimation module.
  • The DistributionEstimator class to estimation module.
  • The root_mean_square() function to estimation module.
  • The GageStudyModel class to model module.
  • The BaseHTMLReprModel class to model module, all Model classes inherit from this class.
  • The CategoricalObservation class to plotter module.
  • The option to stretch the figsize by an amount for AxesFacets and all Chart classes.
  • The MeasurementUncertainty class to estimation module.
  • The property k (coverage factor) to LocationDispersionEstimation class.
  • The options strategy, agreement and possible_dists to GageEstimator class.

Removed

  • The calculation of measurement uncertainties at GageEstimator class.

[1.6.0] - 2025-06-05

Changed

  • Renamed the module templates to precast.
  • Renamed Parameter class to Specification.
  • The groupby option observed to True within Chart and Plotter classes, due to FutureWarnings.
  • LinearModel does not require calling fit method by default after initialization.
  • The attribute name tick_lables to tick_labels at Dodger class.

Added

  • The inclination_displacement() function to montecarlo module.
  • The coaxial() and perpendicular() methods to RandomProcessValue class at montecarlo module.
  • The GageEstimator class to estimation module.
  • The GageRnRModel class to model module.
  • The GageRnRCharts class to precast module.
  • The t_test() function to hypothesis module.
  • Unit tests across many classes.
  • Documentation for plotting guide and numerous examples across many classes.
  • The datasets grnr_adjustment.csv, grnr_layer_thickness.csv and grnr_spc.csv.
  • The option margin to estimate_kernel_density and that this option can also be used in the GaussianKDE and GaussianKDEContour plotters.
  • The option to get the full data range for Estimator lcl and ucl when setting strategy to 'data' and agreement to '1.0' or float('inf').
  • The mask_missing(), mask_ok() methods to Estimator class.
  • The mask_error(), mask_nok() and mask_ok() methods to ProcessEstimator class.
  • The pos_to_ticklabels() method for Dodger class.
  • The Stem class to plotter module.

[1.5.0] - 2025-04-29

Changed

  • LabelAxes draws all labels within the figure and adjusts the subplots. This ensures that all labels are always rendered.

Fixed

  • Term names for VIF and ANOVA tables are now converted correctly even if there are spaces in the factor levels.

[1.4.5] - 2025-04-10

Fixed

  • The images should now be displayed on GH pages and PyPI.

[1.4.4] - 2025-04-10

Fixed

  • Fixed documentation and docstrings to eliminate mkdocs warnings when building documentation.
  • Typing in montecarlo module was not suitable for numpy v2.0.0.

Changed

  • Enhanced README.

[1.4.3] - 2025-04-04

Changed

  • RandomProcessValue requires a Parameter instance instead of nominal and tolerance.

Added

  • Tests for RandomProcessValue class.
  • Examples to docstring of RandomProcessValue class.

[1.4.2] - 2025-04-03

Added

  • Binning class for binning precise values into a specified number of bins.
  • Tests for Binning class.

Removed

  • The functions float_to_bins and precise_to_bin_nominals from montecarlo module. They are now part of Binning class.

[1.4.1] - 2025-04-02

Added

  • The option kind to choose between 'linear' and 'quantile' approach for float_to_bins function in montecarlo module.
  • The function precise_to_bin_nominals in montecarlo module.

[1.4.0] - 2025-04-02

Fixed

  • LinearModel returned wrong parameters at highest_parameters method.
  • Categoical features are now correctly handled by GaussianKDE.
  • LinearModel failed when converting dtype to 'category' in optimize method.
  • LinearModel residual_data method failed for non-numeric index values ​​because they were used for "observation." Now, a separate data series is created.
  • ProcessEstimator n_nok was calculated incorrectly if usl was not present or 0.
  • ProcessEstimator check if lsl and usl are boolean True can lead to incorrectly predictive nok values.

Changed

  • LinearModel method name highest_features to highest_parameters.
  • The option center_points to show_center at CenterLocation plotter, due to consistency.
  • Individual options for mean, median and control_limits at stripes method for ProcessCapabilityAnalysisCharts
  • GaussianKDE and Violine uses now the same call method for plotting.
  • No tuple allowed anywhere where spec_limits must be defined.
  • ProcessEstimator property name "nok_pred" to "nok_fit.

Added

  • The _kw_fill method to SpreadOpacity class.
  • The stripes method for ParameterRelevanceCharts.
  • The option dist when instantiating ProcessCapabilityAnalysisCharts.
  • The property nok_norm to ProcessEstimator class.
  • The p_ks to the returned values of describe method of ProcessEstimator class.
  • Tests for LinearModel highest_parameter, optimize and predict methods.
  • Tests for ProcessEstimator nok_fit and nok_norm property.

[1.3.0] - 2025-02-12

Fixed

  • PairComparisonChart failed when plotting MeanTest, because of n_groups was not set.
  • The method anova() of LinearModel class failed for typ "I" under certain conditions. Automatically changed to "III" for the moment.
  • The function transpose_xy_axes_params() did wrong stuff if axes were shared.

Changed

  • Dependency of matplotlib to >= 3.10.0
  • Method description() of Estimator class now returns a DataFrame instead of a Series.
  • Terminology for "term", "feature" and "parameter" at LinearModel class so that it is used consistently across all methods.

Added

  • Method include() for LinearModel class to re include a parameter.
  • Method optimize() for LinearModel class to optimize the model.
  • Method predict() for LinearModel class to predict the response.
  • Examples for LinearModel class.
  • Example for chart precast.

[1.2.0] - 2025-02-12

Fixed

  • Transpose xy-axes parameters did not work for MultivariateCharts.
  • LinearModel getting captions failed when printing tables.
  • SingleChart had wrong key for getting color at plot method.

Added

  • Option to highlight quantiles in the fill area of Violine and GaussianKDE Plotters.

Changed

  • The name of the Plotter Quantiles to QuantileBoxes
  • QuantileBoxes draws now not overlapping boxes

[1.1.1] - 2025-01-24

Added

  • Option target_on_y for MultivariateChart

[1.1.0] - 2025-01-23

Added

  • The options rows and cols in label method of JointChart used to labell the axes rows and columns.
  • The Estimation class got now the min and max properties.
  • The option exclude added to Estimation describe method, used to exclude variables from the analysis.
  • The describe method of ProcessEstimation returns now also ok and nok as percentage values.
  • A Beeswarm plotter that is comparable to the Jitter plotter. However, with Beeswarm the points are not distributed randomly, but arranged from the inside to the outside.
  • A Quantiles plotter that is comparable to a boxplot. However, quantiles does not draw whiskers but rather just several boxes that show the given number of percent of the observed data.
  • The option to change color, marker, target_on_y, size and width in plot method for SingleChart and JointChart.
  • The option hide_axis for each Plotter class during initialization.
  • The option visible_spines for each Plotter class during initialization.

Changed

  • The class name MultipleVariateChart to MultivariateChart

Removed

  • The option show_density_axis for GaussianKDE. Use hide_axis and visible_spines instead.

[1.0.1] - 2025-01-09

Fixed

  • Calculating Z-score for ProcessEstimator failure when no specification limit is provided.

[1.0.0] - 2025-01-09

Added

  • ProcessEstimator z_transform method and tolerance property.
  • GaussianKDE ignore_feature option for plotting all at same axis base or not.
  • GaussianKDE fill option for filling in the curves.
  • SpreadWidth center calculation option kind.
  • ProcessEstimator Z_lt property for long-term sigma level.
  • ProcessEstimator Z property as sigma level process capability.
  • Confidence interval functions for Cp and Cpk in confidence module.
  • The function estimate_capability_confidence function in estimation module.
  • CapabilityConfidenceIntervall Plotter class.
  • Datasets drop_card.csv and salt_sticks.csv.
  • AxesFacets mosaic option for more flexability in axes arrangement.
  • AxesFacets properties flat and shape.
  • Loess class in estimation module used for plotting loess curves.
  • Plotting styles ggplot2, daspi-dark and seaborn.
  • The module appereance in plotlib used for managing loading and saving styles.
  • Colormaps used at seaborn to the module appereance.

Changed

  • Estimator agreement property gets allways the multiplier of sigma agreement.
  • Chart property name default_kwds changed to kw_default.
  • JointChart possability to add axes independend stripes for mean, median and control_limits
  • Errorbars requires now the amount of groups. This is necessary for bonferroni adjustment.
  • Chart property name axes_facets changed to Axes.
  • Most of the arguments is now only available as a keyword argument in Chart classes.
  • When target_on_y is set to False in SingleChart, the x and y axes params are swapped.
  • The default plotting color is now black and the palette behaves like ggplot2. That means that the colors changes with the number of groups (except for the first group).

Removed

  • The marker option for TransformPlotter.

Fixed

  • ErrorBarsPlotter and JitterPlotter now use the marker option.
  • GaussianKDEPlotter color option now has an effect.
  • JointChart handles and labels now get the stripes stuff from all axes and not only from the last one.
  • Categorical floating features are now plotted as categorical.

[0.5.1] - 2024-11-19

Added

  • GaussianKDEContour Plotter for plotting bivariate distributions as contour lines.
  • kernel_density_estimation_2d function used to calculate bivariate kernel density estimations.
  • Stripe, StripeLine and StripeSpan Plotter for plotting stripes using StripesFacets class or within BlandAltman Plotter.
  • Documentation site

Changed

  • StripesFacets now uses the new StripeLine and StripeSpan Plotters.
  • BlandAltmanPlotter now uses the new StripeLine and StripeSpan Plotters.
  • The target_on_y option is moved from StripesFacets draw method to init.
  • Getting Figure and Axes behavior if not given during initialization of a Plotter class. The plt.gca() function is now used.

Removed

  • Option marker when initializing a Line Plotter. Set the marker if needed when calling the plotter.

Fixed

  • Charts got wrong colors if not given during initialization.

[0.4.0] - 2024-10-24

Added

  • BivariateUnivariateCharts template, provides a set of charts for visualizing the relationship between a target variable and a feature variable.
  • Option for Plotters to change the marker. This can be set during initialization of the Plotter.
  • Option for Charts to change markers, colors and amount of size bins for categorical differentiation.
  • Single feature or target label for JointCharts if all aubplots shares the corressponding axis.

Fixed

  • Categorical feature failed when using floats. Categorical features failed when using floats. Dodging is now always performed, but only with one category if dodge was set to False.

[0.3.0] - 2024-10-09

Added

  • PairComparisonCharts template, provides a set of charts for visualizing the pairwise comparison of two variables.
  • kw_where argument to filter data when calling the plot method of a Chart object. For more information see the documentation pandas.DataFrame.where

Removed

  • BaseTemplate, all template chart classes inherits directly from JointChart.

[0.2.0] - 2024-07-10

Added

  • Option for LinearModel and variance_inflation_factor function to choose if generalized or only straightforward VIFs should be calculated.
  • Option for LinearModel if all categorical features should be one-hot encoded or not.
  • ParameterRelevanceCharts template and BaseTemplate as parent for ResidualsCharts and ParameterRelevanceCharts.
  • anguage dependent strings for sum of squares label, effects label and ParameterRelevanceCharts texts.
  • Option to skip plotting nan values for all TransformPlotters.
  • Property effect_threshold for LinearModel, which returns the effect value for the given alpha risk.
  • Option to skip evaluation of intercept as least significant term for LinearModel.
  • Property design_matrix for LinearModel, which returns the currently encoded design matrix.
  • Option to draw a percentage line when using the Pareto plotter or not.
  • GVIF, its threshold, collinearity decision and method used for the VIF table.

Changed

  • Variance inflation factor function calculates now only values for equal orders.
  • Variance inflation factor function returns a DataFrame instead of a Series
  • Renamed attribute of LinearModel: exclude -> excluded.
  • least_term method of LinearModel when evaluating by effects and some effects are equal, the highest interaction order is returned. Now it also takes into account nan values.
  • eliminate method of LinearModel returns self for method chaining.
  • Effects are now calculated as absolute t-values.
  • Effects table keeps now the original term order (was messed up).

Fixed

  • has_insignificant_term method of LinearModel could not detect minimal terms when the attribute skip_intercept_as_least was set to True.
  • Degrees of freedom for VIF table was calculated wrong.

[0.1.0] - 2024-07-01

Initial release.

Added

  • Character strings in different languages ​​for charts and tables. The language can be set initially and the output adapts automatically. Available languages: German and English.
  • Constants that are available globally throughout the package.

Anova module:

  • LinearModel for running regressions with higher order interactions and analyzing significant terms and their elimination.
  • Tables such as ANOVA, VIF and residuals.

Statistics module:

  • Estimator for calculating statistics such as mean, stdev, skew and excess.
  • ProcessEstimator as an extension of the Estimator class which also takes specification limits into account to calculate Cp and Cpk values.
  • Confidence interval functions for position, spread, proportions and fit lines.
  • Estimation functions for kernel density and distributions
  • Hypothesis tests for normality, homogeneity of variance and independence.

Plotlib module:

  • SingleChart for plotting single Axes charts with different plotters.
  • MultivariateChart for plotting charts with multiple Axes, with the plots divided into the individual Axes depending on the categorical variates.
  • JointChart for plotting multiple axes to which different plotters can be applied.
  • Any type of plotters such as Bar, Violin, KDE, Scatter, Line, Errorbar, Confidence Intervals, Pareto, Bland Altmann, Parallel Coordinate and Jitter.
  • Pre-built Matplotlib stylesheet
  • Classifiers to classify the plots into different colors, markers or marker sizes depending on the categorical variates.
  • Facets for plotting multiple charts in one figure, with the plots divided into the individual Axes depending on the categorical variates. Also for adding stribes to the charts or adding labels, titles