How can we create poll-like plots with confidence intervals like the example attached?

Hi all,

I have a quick question, does anyone know where we can start for this type of chart below?

https://en.wikipedia.org/wiki/2019_Canadian_federal_election#/media/File:Opinion_polling_during_the_pre-campaign_period_of_43rd_Canadian_federal_election.svg

It's technically a regression line, confidence interval and actual data based on time. But graphically it looks so appealing.

I appreciate any clue or advice you could provide.

Thanks

[dpb]

The CurveFitting and/or Statistics TB fit function includes LOESS/LOWESS models although I've not (yet) used the MATLAB implementation.

The Signal Processing TB includes an implementation of the underlying Savitsky-Golay filter.

Generating the plots would be adding features with patch() or the like to shade the intervals -- not terribly difficult but can be time-consuming to build.

Answer 1

[Star Strider]

That would appear to be a scatter plot with mean or median values plotted on it as well (likely using movmean or movmedian), with patch objects providing the shading between the confidence intervals (or something similar).

Example —

t = linspace(1,1000,200);
y = sin(2*pi*t*0.003) + randn(size(t))*0.5;

solidline = movmean(y, 100);
stdev = movstd(y, 100);

figure
scatter(t, y, 5, 'b', 'filled')
hold on
plot(t, solidline, '-r')
patch([t, fliplr(t)], [(solidline-stdev) fliplr(solidline+stdev)], [1 1 1]*0.6, 'FaceAlpha',0.25, 'EdgeColor','none')
hold off
grid

.

[Wolfgang McCormack]

@Star Strider wow you are my hero, thankssssss!

[Star Strider]

Thank you!

As always, my pleasure!