How to average a time series with different length by days?

Hi.

I want to average all the hours from one day into one. So that all the values from one day is averaged together. Any suggestions would be of great help!

- Kristine

My time series looks like this:

Year Month Day Hour Minute Second PM2.5

2009  5  4  22  0  0  15

2009  5  4  23  0  0  20

2009  5  5  0  0  0  16

2009  5  5  1  0  0  16

2009  5  5  2  0  0  16

2009  5  5  3  0  0  16

2009  5  5  4  0  0  17

2009  5  5  5  0  0  17

2009  5  5  6  0  0  21

2009  5  5  7  0  0  17

2009  5  5  8  0  0  20

2009  5  5  9  0  0  18

2009  5  7  15  0  0  13

2009  5  7  16  0  0  17

2009  5  7  17  0  0  18

[Jon]

Not sure what PM2.5 is, but you can convert your dates to datenumbers (see http://www.mathworks.com/help/matlab/ref/datenum.html) and then do the averaging.

DateNumber = datenum(Y,M,D,H,MN,S)

You can use a for loop to find where the day changes and do the averaging within that window. You can also use indexing and logical statements to avoid the for loop.

Answer 1

[Walter Roberson]

Assuming your data is named TS

[unique_rows, ~, idx] = unique(TS(:,1:4),'rows');
daily_av = accumarray(idx(:), TS(:,7), [], @mean);
avg_TS = [unique_rows, daily_av(:)];

[Kristine]

I could not get this code to give me the averaging:/

[Walter Roberson]

[unique_rows, ~, idx] = unique(TS(:,1:3),'rows');

[Kristine]

Thank you very much!!