MATLAB: How to fill NaN’s by averaging previous rows

nantable

I have a table named T2.mat. This table has 12 columns. The last column sometimes has NaN values. I want to fill NaN in this column by the average of columns 9 and 10 in the same rows.
I attached my file.
Thanks in advance

Best Answer

For matrix m
% Create a 10x12 matrix with NaNs in col 12
m = randi(100,10,12);
m(randi(10,1,7),12) = NaN; 
% Replace nans in col 12 with mean of col 9,10
m(isnan(m(:,12)),12) = mean(m(isnan(m(:,12)),[9,10]),2)
For table T
% Create a 10x12 table with NaNs in col 12
m = randi(100,10,12);
m(randi(10,1,7),12) = NaN; 
T = array2table(m); 
T.m12(isnan(T.m12)) = mean([T.m9(isnan(T.m12)), T.m10(isnan(T.m12))],2);
For your data, that will look like, 
T2.tm_m(isnan(T2.tm_m)) = mean([T2.tmax_m(isnan(T2.tm_m)), T2.tmin_m(isnan(T2.tm_m))],2);
If T2.tmin_m or T2.tmax_m contain a NaN, the mean will be a NaN. If 1 of the values are NaN and other is not, you could return that single value using 
T2.tm_m(isnan(T2.tm_m)) = mean(. . .,2,'omitnan');
Related Question