Multiple axis breaks - matlab

Question

I'm trying to create a plot on Matlab with multiple axis breaks (so something like the following):
I've tried using things like breakyaxis and breakaxis from the Matlab File Exchange, but those only allow for one break, not multiple.
Is there any way to implement this?

Answer 1

The NaN (Not a Number) values can be an annoying thing but also a convenient one in some case.
When you plot data, Matlab will leave a blank in place of every data point which has no value (the NaN). So the principle is to insert these NaN between your datasets and tell Matlab to plot the whole lot. Matlab will leave a blank automatically everywhere there is a NaN.
Here is an example, since you didn't supply sample data I first have to define 3 short data sets resembling the ones you have in your figure:
%% // sample data sets
yf = #(x) 2*x+40+randi(7,size(x)) ;
x1 = 57:61 ; y1 = yf(x1) ;
x2 = 72:76 ; y2 = yf(x2) ;
x3 = 80:83 ; y3 = yf(x3) ;
This is an edited answer to take into account the breaks in the Y axis. To be able to call global operations on the datasets I have to regroup them into a cell array or a structure. The struture approach would use loops on the different data sets, while the cell array allow the use of cellfun to compact the code. I chose this approach and use cellfun extensively.
So first step is put all your data sets in a cell array
%% // have to group the data sets in a cell array or structure to implement global operations
xc = { x1 ; x2 ; x3 } ;
yc = { y1 ; y2 ; y3 } ;
Now the heavy part:
%// find the maximum vertical span of the datasets and the total span
maxVal = cellfun(#max,yc) ;
minVal = cellfun(#min,yc) ;
maxYspan = max( maxVal-minVal ) ;
totalSpan = max(maxVal)-min(minVal) ;
%// find a sensible Y value to add between the datasets, not too wide but
%// enough to see a break`
yBreakIncrement = round( totalSpan / 10 ) ; %// adjust that if necessary
yTickIncrement = round( maxYspan /5 ) ; %// adjust that if necessary
%% // rebuild the Y datasets
%// value to substract to each data set to bring them together (including the break space)
setSubstract = [0 ; cumsum( (minVal(2:end)-maxVal(1:end-1))- yBreakIncrement ) ] ;
%// get 3 new data sets brought together
Yall = cellfun(#minus , yc , num2cell(setSubstract) , 'uni',0) ;
%// concatenate the data sets, inserting NaN in the middle
Yall = cellfun( #(a,b) cat(2,a,b) , Yall , repmat({NaN},length(yc),1) , 'uni',0) ;
Yall = cat( 2, Yall{:} ) ;
%// remove the last trailing NaN
Yall(end) = [] ;
%% // Build the Y labels
%// generate ticks that covers each interval
Y_tickpos = cellfun(#colon, num2cell(minVal), repmat({yTickIncrement},length(yc),1) , num2cell(maxVal) , 'uni',0) ;
%// generate the Y labels based the real Y values
Y_labels = cellstr( num2str( cat(2, Y_tickpos{:} ).') ) ; %'// ignore this comment
%// now adjust the actual position
Y_tickpos = cellfun(#minus , Y_tickpos , num2cell(setSubstract) , 'uni',0) ;
Y_tickpos = cat( 2, Y_tickpos{:} ) ;
%% // Build the X labels (and axis)
%// create a continuous index for the X axis
X = 1:length(Yall) ;
X_labels = cellstr( num2str( cat(2, xc{:} ).') ) ; %'// generate the X labels based the X values
X_tickpos = X(~isnan(Yall)) ; %// prepare a vector for the label positions
%% // Display
plot(X,Yall) %// plot as usual
%// Set the labels at the chosen positions
set(gca, 'XTick' , X_tickpos , 'XTickLabel' , X_labels )
set(gca, 'YTick' , Y_tickpos , 'YTickLabel' , Y_labels )
That should give you something like:
Hopefully enough to get you started. Try to adapt the principle to your data.