Calculate difference from a time series
显示 更早的评论
Hello all, i would appreciate your help to solve the following issue (avoiding unnecessary loops if possible). I have a time series of temperature with 4 daily values corresponding in times 0, 6, 12 and 18. I would like to create a new one where i will keep the day and the difference between the value temperature in 12 hr and 6 hr.
My dataset looks like
2007 1 1 0 5
2007 1 1 6 7
2007 1 1 12 14
2007 1 1 18 11
2007 1 2 0 8
2007 1 2 6 9
2007 1 2 12 12
2007 1 2 18 13
where first column is year, second month, third day, fourth time and sixth temperature. The new dataset i want to have will look like
2007 1 1 7
2007 1 2 3
Thank you in advance for your help
采纳的回答
Hi. Clumsy code but should do the trick. Used vectorized code to avoid loops.
% I am assuming your dataset is a matrix
A = [2007 1 1 0 5;
2007 1 1 6 7;
2007 1 1 12 14;
2007 1 1 18 11;
2007 1 2 0 8;
2007 1 2 6 9;
2007 1 2 12 12;
2007 1 2 18 13 ];
% First thing you can do is select the rows of interest.
ii = 2:4:length(A);
jj = 3:4:length(A);
matrix = [A(ii.',:); A(jj.',:)];
% Remove fourth column
% matrix(:,2) = [];
% Sort row in ascending order
matrix = sortrows(matrix);
ll = 1:2:size(matrix, 1);
kk = 2:2:size(matrix, 1);
B = [matrix(kk.',:)];
C = [matrix(ll.',:)];
output = [B(:,1:end-2) B(:, end) - C(:, end)];
更多回答（1 个）
Guillaume
2015-7-8
I would do it like this:
temps = [
2007 1 1 0 5
2007 1 1 6 7
2007 1 1 12 14
2007 1 1 18 11
2007 1 2 0 8
2007 1 2 6 9
2007 1 2 12 12
2007 1 2 18 13];
[ymd, ~, idx] = unique(temps(:, 1:3), 'rows');
tempperday = nan(max(idx), 4);
tempperday(sub2ind(size(tempperday), idx, temps(:, 4)/6 + 1)) = temps(:, 5)
The tempperday array is pretty much the same data as temps but each row is a single day, and the columns are temperature at 0, 6, 8, 12 hours. Therefore to get the difference between 12 and 6:
diffperday = tempperday(:, 3) - tempperday(:, 2)
And to attach the day to it:
newtemps = [ymd diffperday]
类别
在 帮助中心 和 File Exchange 中查找有关 Smoothing and Denoising 的更多信息
Translated by 
