Missing value: Adding new row of NaN in a time series. Need some twist in code

Question

khandu 2013-10-15

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/90235-missing-value-adding-new-row-of-nan-in-a-time-series-need-some-twist-in-code

评论： sixwwwwww 2013-10-15

Dear all,

I want to create some new rows of NaN (lets say NaN(1,:,:)) in gridded time series (e.g., A_matrix (200,40,50)) to fill up some missing months in between. I have identified the missing months using:

[missingvalues,id] = setdiff(actual_period,data_dates);

where id is the index of missing month OR

[locCol, loc] = ismember(actual_period,data_dates);

where loc is an index of logical values for 0=missing month.

Based on the index values I need to add extra rows (NaNs) to A_matrix so that I can interpolate them.

Kindly help me... Khandu

2 个评论
显示无隐藏无

Simon 2013-10-15

Hi!

Do you have some kind of example code to show us?

khandu 2013-10-15

在 MATLAB Online 中打开

HI Simon, Thanks for the response. I have the following code as an example:

    %%A_matrix contains the real observation at time intervals defined by data_period
% it contains missing data at various intervals 
% nmonths refers the months and extra raw needs to be added to match with
% actual_period
A_matrix = rand(122,20,30);
[nmonths, nlat, nlon] = size(A_matrix);
% data period
% convert to decimal year using year and month functions from fnancial
% toolbox
data_period  = datenum(2003,[1:7, 10:35, 38:126],1); % containing missing values at various intervals
data_dates = year(data_period)+month(data_period)/12;
% study period 
actual_period = datenum(2003,1:126,1); %%period defining the study length
actual_dates = year(actual_period)+ month(actual_period)/12;
[locCol, loc] = ismember(actual_dates,data_dates);
% missing time series shall be replace by NaN
missing_dat = NaN(1,nlat, nlon);
% index mising data
[missingvalues,id] = setdiff(actual_dates,data_dates);
% using ismember

From here I m trying to replace the missing timeseries with missing_dat using the index but stuck for sometime.

I hope you can help it.I hope you have the year month function.

Thanks Khandu

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

sixwwwwww 2013-10-15

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/90235-missing-value-adding-new-row-of-nan-in-a-time-series-need-some-twist-in-code#answer_99720

在 MATLAB Online 中打开

Dear Khandu, you can insert NaN at missing location as follows:

A_matrix = rand(122,20,30);
[nmonths, nlat, nlon] = size(A_matrix);
data_period  = datenum(2003,[1:7, 10:35, 38:126],1);
data_dates = year(data_period)+month(data_period)/12; 
actual_period = datenum(2003,1:126,1);
actual_dates = year(actual_period)+ month(actual_period)/12;
missing_dat = NaN(1,nlat, nlon);
[missingvalues,id] = setdiff(actual_dates,data_dates);
ExtraMonths = length(id);
B = zeros(nmonths+ExtraMonths, nlat, nlon);
A_index = 1;
for i = 1:nmonths+ExtraMonths
    if sum(ismember(id, i)) == 1
        B(i, :, :) = missing_dat;
    else
        B(i, :, :) = A_matrix(A_index, :, :);
        A_index = A_index + 1;
    end
end