Corrcoef creating NaNs from datasets with no NaNs

I have been trying to analyse correlations between temperature and pressure fields or a temperature field and a time series and for some reason even though neither the temperature or pressure datasets have any NaNs in them corrcoef is producing NaNs.

What reasons are there for this to happen?

The NaNs that are being produced are propagating through my program and causing problems with the output.

Answer 1

[Youssef Khmou]

hi, because your data contain similar values, here is an example :

 corrcoef([ones(4,1) ones(4,1)])

Answer 2

[Sean de Wolski]

corrcoef(zeros(10,1),zeros(10,1))

They'll occur from 0/0 or inf/inf

[Rhiannon]

The values in both datasets are quite different to each other. The temperature is around 250-300, pressure 950-1050 and the time series is -10-20

[Sean de Wolski]

It's hard to say without seeing the data. You could run the following and then run your code:

dbstop if naninf

( http://www.mathworks.com/help/releases/R2013b/matlab/ref/dbstop.html )

Which will stop with the debugger when the NaN is calculated and you can inspect what's going on.