removing entire row in cell array if first column is a NaN

My data matrix (after reading into matlab with xlsread) contains many rows of string and numerical data, and at the end, there sometimes exists several rows of "NaN"s

I want to delete these specific rows but can't seem to do it properly.

I've tried q=cellfun(@(x) all(isnan(x)),myMatrix); myMatrix(q); but the result is no longer a cell

I've tried findNan=isnan(myMatrix(:,1)) to find the index of the row containing the NaN's to kill it, but it gives an error: ??? Undefined function or method 'isnan' for input arguments of type 'cell'.

Any assistance is appreciated! Thanks :)

[Walter Roberson]

xlsread() can return 3 matrices, one for numeric, one for text, one for raw. Which one are you working with? To get the mix of numeric and text you must be working with raw ?

[David C]

i'm working with the RAW matrix ... this is a must for me at the moment

Answer 1

[Walter Roberson]

myMatrix( cellfun( @(C) isnumeric(C) && isnan(C), myMatrix(:,1) ), :) = [];

[David C]

Walter, thanks that seemed to it, would you mind explaining the cellfun part of the code a little bit?

I dont understand the isnumeric && isnan part of it

Thank you kindly

[Walter Roberson]

You have cells that contain strings; strings fail the isnumeric() test, and the && operator does not go on to the second test if the first test fails. isnan() is thus only applied to numeric values, and you only have one numeric value per cell so isnan() will return a scalar value, so you do not need any(). If you left off the isnumeric() test, then isnan() would be applied to the strings; it is well defined on strings but it returns one value per character.

An alternate coding should be

myMatrix( cellfun( @(C) any(isnan(C)), myMatrix(:,1) ), :) = [];

On the other hand, if somehow the matrix could contain other data types in the cells, then the isnumeric() version would be safer. Safer yet would be

myMatrix( cellfun( @(C) isnumeric(C) && any(isnan(C(:))), myMatrix(:,1) ), :) = [];

which allows for the possibility that a numeric array might have slipped in

[David C]

super!! My data is consistent, and in actuality, I only have to search down the first col of my cell array which is an array of strings (which is what this code is actually doing correct? the myMatrix(:,1) part of it?)

then it searches that array and performs the isnumeric and isnan functions on it till it gets a true "1" value, and that row's index is returned by cellfun, which in turn is made into a blank [] am I understanding that correctly?

If so, since I know my first column in the cell array will always be strings, I technically don't need the isnumeric(C) part of it right?

Thanks again!

Answer 2

[the cyclist]

Does this do what you want?

% Some made-up data
c = num2cell([1 1; 2 2; NaN NaN]);

% Find the rows with all NaN
indexToRowToDrop  = all(arrayfun(@(x)isnan(x{:}),c),2)

% Drop the bad rows
c(indexToRowToDrop,:) = [];