Filling in secondly intervals between 15:30:00 to 22:00:00 (require more efficient method)

I have data in the form of

Hi all I have prices which change every 15 seconds throughout the whole day.

I want to make the the price secondly rather than every 15 seconds. This is so that I can compare it to another second-second data.

Here is a sample of my data:

'26/01/2012 15:29:39' 168.008200000000

'26/01/2012 15:29:54' 168.042200000000

'26/01/2012 15:30:09' 168.022300000000

'26/01/2012 15:30:24' 167.964000000000

'26/01/2012 15:30:39' 167.968800000000

'26/01/2012 15:30:54' 167.964000000000

'26/01/2012 15:31:09' 167.973700000000

'26/01/2012 15:31:24' 167.954700000000

Like I said I want the data to be second-second from 15:30:00 to 22:00:00. So for example:

From the period of 15:30:00 - 15:30:09 we would maintain a price of 168.0422

Then from the period of 15:30:09 to 15:30:24 we would maintain a price of 168.0223 secondly.

I hope I have explained this well and I look forward to all replies.

The "raw" data is calculated 15 seconds and even go beyond 22:00 and is irregular.

"" The program would need to take any data in the form of below and then just give secondly prices from 15:30:00 - 22:00:00.

Thanks

[Mate 2u]

I will be reading in 26/10/2012 for this but the program needs to be compatiable with different dates

[Jan]

Please do not post multiple threads for one problem. See: http://www.mathworks.com/matlabcentral/answers/34419-filling-in-secondly-time-intervals-between-15-30-00-to-22-00-00

Did you try my answer of your first thread before deleting it?

[Mate 2u]

Hi Jan, Yes I did try your answer. Thanks so much. Unfortunately it was not working.

[Jonathan Sullivan]

It usually helps the community to have a record of what you have tried already.

[Thomas]

Another thought: why do you just reduce the other value you are comparing this data set to, to be every 15 seconds..

[Mate 2u]

Hi Jonathan, Yes I do understand I wont make that mistake again.

Thomas. I have two time series which have the same date formats. One is prices of a particular asset which change many times in a second. I am comparing it to another asset which is "priced" every 15 seconds. I need it every second so that I can decide "trade" decisions every second.

Thanks.

Answer 1

[Jan]

I try it again. I have no access to Matlab currently. Please test this an post, where it fails together with the error message or an explanation of the difference between your expectations and the results.

C = {'26/01/2012 15:29:39', 168.008200000000; ...
     '26/01/2012 15:29:54', 168.042200000000; ...
     '26/01/2012 15:30:09', 168.022300000000; ...
     '26/01/2012 15:30:24', 167.964000000000; ...
     '26/01/2012 15:30:39', 167.968800000000; ...
     '26/01/2012 15:30:54', 167.964000000000; ...
     '26/01/2012 15:31:09', 167.973700000000; ...
     '26/01/2012 15:31:24', 167.954700000000};

Time     = round(datenum(C(:, 1), 'dd/mm/yyyy HH:MM:SS') * 86400)
allTime  = Time(1):Time(end);
index    = cumsum(ismember(allTime, Time));
Value    = cat(1, C{:, 2});
allValue = Value(index);

[Mate 2u]

Hi there, I tested this with my data. I am getting the allValue and allTime as array of size 23406 where it should be 23401 (23401 seconds in 15:30:00 - 22:00:00).

[Mate 2u]

but it is fine good work I just delete the end points

[Matt Tearle]

@Jan: beautiful!

@Mate 2u: Note that Jan just used the first and last points of Time to make the fine-grain time vector. If you want specific values (15:30:00 to 22:00:00) you could just hard-code those in. But also note that Jan's solution is using date numbers scaled up by 86400 (= the number of seconds in a day) so each integer value is a second. Hence:

startTime = datenum('26/02/2012 15:30:00' 'dd/mm/yyyy HH:MM:SS')*86400;

endTime = datenum('26/02/2012 22:00:00' 'dd/mm/yyyy HH:MM:SS')*86400;

allTime = startTime:endTime;

You could even figure out the start and end times automatically, if you had some rationale behind how you chose them.