Replacing nans and greater than

I have a file with 3 columns. I need to change all the NaNs in the second column of the cell array to empty cells, and the corresponding row in column 3 with number 4. I also need to do the same for numbers greater than 0.2, changing row 3 to a 5.

I tried something like this:

if database(:,2) = is a nan then it will change to []
    database(:,3) = 4

if database(:,2) > 0.2 
    database(:,3) = 5

[James Tursa]

What do you get when you type:

whos database

Answer 1

[Geoff Hayes]

Nicole - the easiest (but not necessarily most efficient) way is to just iterate over each element of the second column and apply your conditions, using the isnan function. Something like

 numRows = size(database,1);
 for k=1:numRows
     if isnan(database{k,2})
         database{k,2} = [];
         database{k,3} = 4;
     elseif database{k,2} > 0.2
         database{k,3} = 5;
     end
 end

Note how the {} braces are used to access (get and set) the elements in the cell array.

[Nicole Scerri]

This is the error i get when I run this:

??? Cell contents reference from a non-cell array object.

[Nicole Scerri]

Thanks you :) It worked when I took off the curly brackets!!

 numRows = size(database,1);
 for i=1:numRows
     if isnan(database(i,2))             %missiong data
         database(i,3) = 4;
     elseif database(i,2) > maxLevel     %high range
         database(i,3) = 5;

Answer 2

[Walter Roberson]

nanpos = cellfun(@(C) any(isnan(C)), database(:,2));
database(nanpos,2) = {};
database(nanpos,3) = {4};
pos02 = cellfun(@(C) any(C>0.2), database(:,2));
database(pos02,3) = {5};

You need to decide whether database is a numeric array or a cell array. If it is a numeric array then you cannot put in empty cells: every entry in a numeric array must be a (possibly NaN) numeric entry. If the database is a cell array, then you cannot put a number in a column: you have to put a cell containing the number.

For the 0.2 condition, notice that I used any(C>0.2) and not C>0.2 directly. We just changed some of the entries to the empty cell a moment before, so we don't know how many numbers there are in the given cell. There might be 0, there might be 1, there might be... well, you didn't define: if you are working with a numeric array then of course there is only one entry but then you would not be able to replace the entries with emptiness. So to make the question work we have to assume that each entry is a cell, and when we do that, it is no longer safe to assume that each cell will be either empty or a scalar. Even if you can guarantee that to be the case, cells of empty array or scalars, then if you use just C>0.2 then if C is empty then []>0.2 is defined as [], but you can't return the [] without changing to 'Uniform', 0, and then you would have to post-process the resulting cells. Easier all around to use any() as that is well defined for empty vectors.

[Nicole Scerri]

When it run that, it give me this error:

??? Error using ==> cellfun Input #2 expected to be a cell array, was double instead.

It's a table like this:

datenum     level       flag
            2.1          0
            NaN          0

[Walter Roberson]

When you say it is a table, do you mean that it was created with table() ? If so then see http://www.mathworks.com/help/matlab/matlab_prog/clean-messy-and-missing-data-in-tables.html