Box Plot with Whiskers Plotted at Fixed Percentiles

This script uses boxplot but plots the whiskers for a specified percentile.

Description

Normally boxplot plots the whiskers with a maximum length based on the distance between the 25th and 75th percentile. This script uses boxplot but plots the whiskers for a specified percentile and only plots the outliers beyond the new whiskers.

Note that the percentile is defined as the data point just beyond the calculated percentile. For example, if the 98th percentile lies between the 4th and 5th highest data points, then the 4th highest data point will be considered the 98th percentile, the whisker will extend to the 4th highest, and only the 4 highest values will be plotted as outliers.

Acknowledgements

The author wishes to acknowledge the following in the creation of this submission:
Quartile & Percentile Calculation

• Required Products: Statistics Toolbox
• MATLAB release: MATLAB 7.7 (R2008b)

[Peter]

unfortunately function doesn't work for me. Seems to include non defined functions as 'nanfree',...

[Stephen]

This is a useful program.

I found that boxplot_percentwhisk does not handle negative-valued outliers correctly; all negative outliers are automatically cut off.

I removed the code from lines 91 - 94:
    outliers = get(OUT(length(group_list)-i+1),'ydata');
    outlier_index = find(outliers == data_highpercent(i));
    outliers(1:outlier_index) = data_highpercent(i);
    set(OUT(length(group_list)-i+1),'ydata',outliers);

and replaced it with the following code to fix the problem:
    outliers_y = get(OUT(length(group_list)-i+1),'ydata', 'xdata');
    outliers_x = get(OUT(length(group_list)-i+1),'xdata');
    new_outlier_index = ((outliers_y < data_lowpercent(i)) | (outliers_y > data_highpercent(i)));
    outliers_y = outliers_y(new_outlier_index); % Find the outliers that
    outliers_x = outliers_x(new_outlier_index);
    set(OUT(length(group_list)-i+1),'ydata',outliers_y, 'xdata',outliers_x);

[Stephen]

Sorry, my bug fix included an error.

Lines 91 - 94 in boxplot_percentwhisk should be removed
    outliers = get(OUT(length(group_list)-i+1),'ydata');
    outlier_index = find(outliers == data_highpercent(i));
    outliers(1:outlier_index) = data_highpercent(i);
    set(OUT(length(group_list)-i+1),'ydata',outliers);

Those lines should be replaced with (corrected version):
    outliers_y = get(OUT(length(group_list)-i+1),'ydata');
    outliers_x = get(OUT(length(group_list)-i+1),'xdata');
    new_outlier_index = ((outliers_y < data_lowpercent(i)) | (outliers_y > data_highpercent(i)));
    outliers_y = outliers_y(new_outlier_index);
    outliers_x = outliers_x(new_outlier_index);
    set(OUT(length(group_list)-i+1),'ydata',outliers_y, 'xdata',outliers_x);