Line of Best Fit for a time series plot

I would like to know how I can use the polyfit function to fit a line to my time series plot.

ts_MinJan = timeseries(Jan_Min,1:41);
ts_MinJan.Name = 'Janurary Min';
ts_MinJan.TimeInfo.Units = 'Year';
subplot(3,4,1), plot(ts_MinJan)
ts_MinFeb = timeseries(Feb_Min,1:41);
ts_MinFeb.Name = 'February Min';
ts_MinFeb.TimeInfo.Units = 'Year';
subplot(3,4,2), plot(ts_MinFeb)
ts_MinMar = timeseries(Mar_Min,1:41);
ts_MinMar.Name = 'March Min';
ts_MinMar.TimeInfo.Units = 'Year';
subplot(3,4,3), plot(ts_MinMar)
ts_MinApr = timeseries(Apr_Min,1:41);
ts_MinApr.Name = 'April Min';
ts_MinApr.TimeInfo.Units = 'Year';
subplot(3,4,4), plot(ts_MinApr)
ts_MinMay = timeseries(May_Min,1:41);
ts_MinMay.Name = 'May Min';
ts_MinMay.TimeInfo.Units = 'Year';
subplot(3,4,5), plot(ts_MinMay)
ts_MinJun = timeseries(Jun_Min,1:41);
ts_MinJun.Name = 'June Min';
ts_MinJun.TimeInfo.Units = 'Year';
subplot(3,4,6), plot(ts_MinJun)
ts_MinJul = timeseries(Jul_Min,1:41);
ts_MinJul.Name = 'July Min';
ts_MinJul.TimeInfo.Units = 'Year';
subplot(3,4,7), plot(ts_MinJul)
ts_MinAug = timeseries(Aug_Min,1:41);
ts_MinAug.Name = 'August Min';
ts_MinAug.TimeInfo.Units = 'Year';
subplot(3,4,8), plot(ts_MinAug)
ts_MinSep = timeseries(Sep_Min,1:41);
ts_MinSep.Name = 'September Min';
ts_MinSep.TimeInfo.Units = 'Year';
subplot(3,4,9), plot(ts_MinSep)
ts_MinOct = timeseries(Oct_Min,1:41);
ts_MinOct.Name = 'October Min';
ts_MinOct.TimeInfo.Units = 'Year';
subplot(3,4,10), plot(ts_MinOct)
ts_MinNov = timeseries(Nov_Min,1:41);
ts_MinNov.Name = 'November Min';
ts_MinNov.TimeInfo.Units = 'Year';
subplot(3,4,11), plot(ts_MinNov)
ts_MinDec = timeseries(Dec_Min,1:41);
ts_MinDec.Name = 'December Min';
ts_MinDec.TimeInfo.Units = 'Year';
subplot(3,4,12), plot(ts_MinDec)

Answer 1

[Mathieu NOE]

hello

I am not working that much with timeseries , but IMHO, you could do the polyfit first on the raw data , otherwise you can still access the data from a timeseries to do the fit then you have to create a new timeserie for the fitted curve to overlay on the plot

demo :

n = 10;
t = (0:n-1)';
x = 1+t/n + 0.1*randn(n,1);

ts1 = timeseries(x,t);

ts1.Name = 'Daily Count';
ts1.TimeInfo.Units = 'days';
ts1.TimeInfo.StartDate = '01-Jan-2011';     % Set start date.
ts1.TimeInfo.Format = 'mmm dd, yy';       % Set format for display on x-axis.
ts1.Time = ts1.Time - ts1.Time(1);        % Express time relative to the start date.

plot(ts1)

% polyfit
% [p,S,mu] = polyfit(t,x,1);  % polyfit on raw data 
[p,S,mu] = polyfit(t,ts1.Data,1); % polyfit on ts1 data 
[xfit,delta] = polyval(p,t,S,mu);

ts2 = timeseries(xfit,t);
ts2.TimeInfo.Units = 'days';
ts2.TimeInfo.StartDate = '01-Jan-2011';     % Set start date.
ts2.TimeInfo.Format = 'mmm dd, yy';       % Set format for display on x-axis.
ts2.Time = ts2.Time - ts2.Time(1);        % Express time relative to the start date.

% Plot fit line: 
hold on
plot(ts2)

Answer 2

[Peter Perkins]

There are a bunch of things you might mean by this, but here's one thing you might do.

% set up random data
Time = datetime(1983:2023,1,1,Format="uuuu");
tt = array2timetable(rand(41,12),RowTimes=Time,VariableNames=["Jan" "Feb" "Mar" "Apr" "May" "Jun" "Jul" "Aug" "Sep" "Oct" "Nov" "Dec"]);

% fit a quadratic to each month's data, plot along with the raw data
t = 1:length(Time);
tt2 = varfun(@(x) polyval(polyfit(t,x,2),t)',tt);
tt2.Properties.VariableNames = tt.Properties.VariableNames;
stackedplot(tt,tt2)