dates missing in time series - how to remove discontinuity

I am plotting some daily time series data and there are about 60 contiguous days missing from the middle of the data set. When i plot, I am getting this ugly discontinuity where there are missing days and I can't, for the life of me, figure out how to close this "gap." I'm using Matlab 2012b on Mac OS X. Thanks in advance.

my code:

figure
plot(dates,Y,'linewidth',1)
set(gca,'XTick',dates)
datetick('x','mm-yyyy')
xlabel('Date')

[Iain]

I suspect dates has some NaN in it. If you:

t = isnan(dates); dates(t) = []; Y(t) = [];

Then you ought to get the plot without the gap.

It may make more sense to interpolate to get the missing dates (if you have valid Y data)

[John]

There are no NANs in the data, I've already removed them. The issue is that dates are on the x-axis and the missing date range just gets skipped and there is a gap with a straight line going from the last data point before the gap to the first data point after the gap.

I've tried both having NaNs and no NaNs, same result. Any ideas?

Answer 1

[José-Luis]

What do you dates look like? If it is something like 800415 (ddmmyy) then there will be a gap at the end of every month and a larger one at the end of every year. If that is the case an idea is to use Matlab day number.

 doc datenum

In the example above:

 serial_date = datenum(800415,'ddmmyy');

The function works with arrays and for plenty of date formats. You could then

plot(serial_data,your_data);

[John]

Yes, I am using datenums. just before the code posted in my question above is a line that is:

dates = datenum(num2str(dates),'yyyymmdd');

I still get the gap. It's basically like matlab knows there are dates missing and just leaves the gap to preserve the distance between xticks. I'm sure things would look strange if you had daily data and it plotted the xaxis like:

    |        |     |        |        |
   2005     2006  2007     2008     2009

instead of:

    |        |        |        |        |
   2005     2006     2007     2008     2009

But the consequence is gaps in the plotted data for the latter method (which is what I'm seeing and can't seem to get the first method).

[José-Luis]

Not entirely sure what you want to achieve. But if you want to get rid of the gap, you could use a continuous date series and fill in with NaN's where there are no data.

Sort of like this.