Using the misdata function on two-dimensional arrays

Dear UserCommunity I need help using the misdata function on two-dimensional arrays with missing data. I can get it to work on vectors of data but it doesn't converge (even over a weekend!!!) on two-dimensional arrays. The Unscrambler missing data algorithm works in a few seconds so I must be doing something wrong. Can anybody post some working source code that uses misdata on a two-dimensional array. Thanks

Answer 1

[Rajiv Singh]

Can you post an example? By "two dimensional" array of data, do you mean data containing two output signals?

Answer 2

[Patrick Jackman]

Sorry I mean a two-dimensional matrix with n-columns of sensor readings and m-time increments. Sometimes a sensor loses communication for a day or even a few days so the "missing" readings must be estimated. I was hoping that the misdata function could estimate the missing data by contructing a model of the silent sensor(s) from those that are live. Thanks

[Rajiv Singh]

Hi Patrick,

The syntax will be as follows:

x = iddata(your_2d_data, [], 1);

x2 = misdata(x, 20, 1e-3);

newdata = x2.y;

Whether this delivers what you expected or not is a different matter; it depends upon the input parameters (maxiter, tolerance values) and the data characteristics. Note that misdata assumes an underlying model structure that can produce the data you have. The default choice of the model structure may not be suitable for your data. See the alternative syntax misdata(data, model) wherein you specify the model that should be used for interpolation.

Answer 3

[Sean de Wolski]

I've never heard of the misdata function. What are the missing data points right now? Are they NaNs? If they are, try John's D'Errico's inpaint_nans on the FEX.

http://www.mathworks.com/matlabcentral/fileexchange/4551-inpaintnans