plotDriftStatus

Visualize p-values and confidence intervals

Syntax

plotDriftStatus(DDiagnostics)

plotDriftStatus(DDiagnostics,Variables=variables)

plotDriftStatus(ax,___)

EB = plotDriftStatus(___)

[EB,CL] = plotDriftStatus(___)

Description

plotDriftStatus(DDiagnostics) plots the estimated p-value for the permutation test for each variable specified for drift detection in the call to detectdrift, as well as the confidence interval for each estimation, using error bars. It also plots the warning and drift thresholds and color-codes the p-values with their confidence intervals according to their drift status.

If you set the value of EstimatePValues to false in the call to detectdrift, then plotDriftStatus does not generate a plot and returns a warning.

plotDriftStatus(DDiagnostics,Variables=variables) plots the drift status for the variables specified by variables.

plotDriftStatus(ax,___) plots into axes ax instead of gca.

EB = plotDriftStatus(___) creates an error bar plot and returns an array of ErrorBar objects in EB. Use this to inspect and adjust the properties of the error bars. To learn more about the properties of the ErrorBar object, see ErrorBar Properties.

[EB,CL] = plotDriftStatus(___) creates an error bar plot and returns two arrays of graphics objects, EB and CL.

EB is an array of ErrorBar objects and CL is an array of ConstantLine objects. Use CL to inspect and adjust the properties of the lines. To learn more about the ConstantLine object properties, see ConstantLine Properties.

Examples

Plot Drift Status for All Variables

Generate baseline and target data with three variables, where the distribution parameters of the second and third variables change for target data.

rng('default') % For reproducibility
baseline = [normrnd(0,1,100,1),wblrnd(1.1,1,100,1),betarnd(1,2,100,1)];
target = [normrnd(0,1,100,1),wblrnd(1.2,2,100,1),betarnd(1.7,2.8,100,1)];

Perform permutation testing for all variables to check for any drift between the baseline and target data.

DDiagnostics = detectdrift(baseline,target)

DDiagnostics = 
  DriftDiagnostics

              VariableNames: ["x1"    "x2"    "x3"]
       CategoricalVariables: []
                DriftStatus: ["Stable"    "Drift"    "Warning"]
                    PValues: [0.3850 0.0050 0.0910]
        ConfidenceIntervals: [2x3 double]
    MultipleTestDriftStatus: "Drift"
             DriftThreshold: 0.0500
           WarningThreshold: 0.1000


  Properties, Methods

Display the 95% confidence intervals for the estimated p-values.

DDiagnostics.ConfidenceIntervals

ans = 2×3

    0.3547    0.0016    0.0739
    0.4160    0.0116    0.1106

Plot the drift status for all three variables.

plotDriftStatus(DDiagnostics)

plotDriftStatus plots the confidence intervals for the estimated p-values, using error bars. It also compares the confidence bounds against the drift and warning thresholds and indicates their drift status using different colors. The lower confidence bound of the p-value for the first variable is higher than the warning threshold, hence the drift status for the first variable is stable, and is indicated by the color blue. The lower confidence bound of the p-value for the third variable is lower than the warning threshold, but higher than the drift threshold, hence the drift status for the third variable is warning, and is indicated by the color yellow. The upper confidence bound of the p-value for the second variable is lower than the drift threshold, hence the drift status for the third variable is drift and is indicated by the color orange.

Plot Drift Status for Specified Variables

Load the sample data.

load humanactivity

For details on the data set, enter Description at the command line.

Assign the first 250 observations as baseline data and next 250 as target data for the first 15 variables.

baseline = feat(1:250,1:15);
target = feat(251:500,1:15);

Test for drift on all variables.

DDiagnostics = detectdrift(baseline,target)

DDiagnostics = 
  DriftDiagnostics

              VariableNames: ["x1"    "x2"    "x3"    "x4"    "x5"    ...    ]
       CategoricalVariables: []
                DriftStatus: ["Drift"    "Drift"    "Drift"    ...    ]
                    PValues: [1.0000e-03 1.0000e-03 1.0000e-03 ... ]
        ConfidenceIntervals: [2x15 double]
    MultipleTestDriftStatus: "Drift"
             DriftThreshold: 0.0500
           WarningThreshold: 0.1000


  Properties, Methods

Display the 95% confidence intervals for the p-values for variables 10 to 15.

DDiagnostics.ConfidenceIntervals(:,10:15)

ans = 2×6

    0.0000    0.4646    0.2225    0.0000    0.5427    0.0766
    0.0056    0.5275    0.2770    0.0056    0.6049    0.1138

Plot drift status for variables 10 to 15.

plotDriftStatus(DDiagnostics,Variables=(10:15))

Change Error Bar Color on Drift Status Plot

Load the sample data.

load humanactivity

For details on the data set, enter Description at the command line.

Assign the first 250 observations as baseline data and next 250 as target data for the first 15 features.

baseline = feat(1:250,1:15);
target = feat(251:500,1:15);

Test for drift on all variables.

DDiagnostics = detectdrift(baseline,target);

Plot drift status for all variables.

[EB,CL] = plotDriftStatus(DDiagnostics)

EB = 
  3x1 ErrorBar array:

  ErrorBar    (Stable)
  ErrorBar    (Warning)
  ErrorBar    (Drift)

CL = 
  2x1 ConstantLine array:

  ConstantLine
  ConstantLine

EB is an array of ErrorBar objects and CL is an array of ConstantLine objects. You can change the appearance of the plot by accessing the properties of these objects. Change the color of the error bars and markers for status'Stable' to green and the color of the drift threshold line, the error bars, and markers for status 'Drift' to magenta.

EB(1).Color = [0 1 0];
EB(1).MarkerFaceColor = [0 1 0];
EB(1).MarkerEdgeColor = [0 1 0];
EB(3).Color = [1 0 1];
EB(3).MarkerFaceColor = [1 0 1];
EB(3).MarkerEdgeColor = [1 0 1];
CL(2).Color = [1 0 1];

You can also access other properties by double-clicking on EB or CL in the Workspace and adjust them using the Property Inspector.

Input Arguments

DDiagnostics — Diagnostics of the permutation testing for drift detection

DriftDiagnostics object

Diagnostics of the permutation testing for drift detection, specified as a DriftDiagnostics object returned by detectdrift.

variables — List of variables

string array | cell array of character vectors | integer indices

List of variables for which to plot the drift status, specified as a string array, a cell array of character vectors, or a list of integer indices.

Example: Variables=["x1","x3"]

Example: Variables=(1,3)

Data Types: single | double | char | string

ax — Axes to plot into

Axes object | UIAxes object

Axes for plotDriftStatus to plot into, specified as an Axes or UIAxes object. If you do not specify ax, then plotDriftStatus creates the plot using the current axes. For more information on creating an axes object, see axes and uiaxes.

Output Arguments

EB — Error bars showing the confidence intervals

3-by-1 array of ErrorBar objects

Error bars showing the confidence intervals for the estimated p-values in the plot, returned as a 3-by-1 array of ErrorBar objects. Use EB to inspect and adjust the properties of the error bars. To learn more about the properties of the ErrorBar object, see ErrorBar Properties.

CL — Lines showing the threshold values

2-by-1 array of ConstantLine objects

Lines showing the drift and warning threshold values in the plot, returned as a 2-by-1 array of ConstantLine objects. Use CL to inspect and adjust the properties of the error bars.

Version History

Introduced in R2022a