How can I delete the extra matrices which contain NaN and 0 values?

1 次查看（过去 30 天）

Szabó-Takács Beáta 2015-4-28

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/213941-how-can-i-delete-the-extra-matrices-which-contain-nan-and-0-values

评论： Szabó-Takács Beáta 2015-4-28

Dear All, I have a 30 year dataset (1961.02.1-1990.11.30) in cell array which contains 358 matrices where the data are stored longitude x latitude x days. I had to convert this cell to 4D matrix for calculation. I carried out this by the following way:

maxSize = max(cellfun(@(x) size(x,3),out_m));
f = @(x) cat(3, x, nan(size(x,1),size(x,2),maxSize-size(x,3))); 
out_m2 = cellfun(f,out_m,'UniformOutput',false);
PL = bsxfun(@times, out_m2, l);
 PL2=reshape(num2cell(PL,[1 2 3]),[],1);
P_scen=cat(3,PL2{:});

Thust every month has 31 days. I would like to write the P_scen into netcdf file. The values are stored longitude x latitude x time. Its time length is 11098 due to the conversation. The time has to be determined in netcdf file by:

T(:,1)=[datenum(1961,2,1,0,0,0):datenum(1990,11,30)];
time(:,1)=T(:,1)-datenum(1950,1,1);

So this time length is 10895. How can I delete the extra days which contain NaN and 0 values? Thank you for your help in advance!

2 个评论
显示无隐藏无

Jan 2015-4-28

在 MATLAB Online 中打开

Note: Nicer:

maxSize = max(cellfun('size', out_m, 3))

Szabó-Takács Beáta 2015-4-28

在 MATLAB Online 中打开

Meantime came to my mind a solution. This is the following:

ARR = P_scen(:,:,:);
a = datenum([1961 2 1; 1990 11 30]);
yy=[1961 1 1; 1961 2 1; 1961 3 1;...; 1990 12 1];
[yyyy,mm,dd] = datevec((a(1):a(2))');
n=numel(dd);
ii = cumsum(ismember([yyyy,mm,dd],yy,'rows'));
pr = accumarray(ii,(1:n)',[],@(x){ARR(:,:,x)});
Pr = cat(3,pr{:});