# Average values with nan value

6 views (last 30 days)
Commented: Adam Danz on 22 Mar 2023
Hi
I have a matrix with 3 dimentions; i,j,t(time).
I try to calculate mean value for each time step but the matter is we have some nan-values. I have four basins and decide to calculate avrage for each basin seperately. It doesnot work properly
............................................
for t= 1:length(time)
if j>85
U1(t)=nanmean(Uwind,'all')
else if j>65 & j<85
U2(t)=nanmean(Uwind,'all')
else if j>25 & j<65
U3(t)=nanmean(Uwind,'all')
else
U4(t)=nanmean(Uwind,'all')
end
end
end
end

Antoni Garcia-Herreros on 16 Mar 2023
Not exacty sure what you mean by U1(t). I you want to calculate the mean (excluding NaNs) for each timestep for all i and a specified range of j this should work.
U4=nanmean(Uwind(:,1:25,:),[1,2]);
U3=nanmean(Uwind(:,26:64,:),[1,2]);
U2=nanmean(Uwind(:,65:84,:),[1,2]);
U1=nanmean(Uwind(:,65:end,:),[1,2]);
Adam Danz on 22 Mar 2023
Instead of nanmean, use mean with the omitnan flag.
For example,
U4 = mean(Uwind(:,1:25,:),[1,2],'omitnan');
mean has more capabilities such as supporting tall arrays, GPU arrays, and code generation, works better with tables, etc. nanmean is discouraged starting in MATLAB R2020b.

Cameron on 16 Mar 2023
myArray = nan(8,1); %array of nan values
indx = 1:2:7; %index to populate that array
myArray(indx,1) = rand(length(indx),1) %have some nan and some numbers
myArray = 8×1
0.1698 NaN 0.0233 NaN 0.8454 NaN 0.6111 NaN
mean(myArray(~isnan(myArray))) %take the average of only the numbers
ans = 0.4124