explain the logic please

Hi

Here is a small piece of code (3rd party)

*************************************************

function A = TR1_4_CountTime(DCA,Time)
% calculate time differences between each row
for i=1:length(DCA)-1
   A(i) = 0 ;
   for j=DCA(i)+1:DCA(i+1)
       A(i) = A(i) + Count(Time{j},Time{j-1}) ;
   end
end

end

function s = Count(time1,time2)

if ~isequal(time1(1:2),time2(1:2)) % if data are not in the same day,
                                   % set default time gap as 1 millisecond
    t1=datenum(time1(find(time1==',')+1:end),'HH:MM:SS')+1;
    t2=datenum(time2(find(time2==',')+1:end),'HH:MM:SS');
    s=round((t1-t2)*24*60); % for minute data
    return ;
else              % if data are in the same day, calculate time difference
    t1=datenum(time1(find(time1==',')+1:end),'HH:MM:SS');
    t2=datenum(time2(find(time2==',')+1:end),'HH:MM:SS');
    s=round((t1-t2)*24*60); % for minute data
end
end

************************************************************

please can somebody explain what the author is trying to do ?

Bob M

Answer 1

[Walter Roberson]

time1 and time2 are character vectors in which day number are represented as the first two characters . At some point in the character vectors, there is a comma, which is immediately followed by characters representing hours (24 hour), minutes, and seconds.

The code extracts the hours, minutes, and seconds of the two times, and finds the difference and converts it to minutes. If the two times were on different day numbers, then then 1 day is added -- even if the dates were multiple days apart.

I would recommend replacing the whole thing with converting to datetime objects and taking round(minutes()) of the difference of the datetime objects.

Answer 2

[Bob Matthews]

Thanks Walter...............

My problem is with the output I see - for example

*****************************

T_EXT PEXT T_DCC PDCC PDCC* OSVEXT TMVEXT T R_DC sub_NDC USV

01/07/2016,00:32:0 102.924 04/07/2016,21:26:0 102.146 102.1521 -2.5153 -3.49643 13 0

06/07/2016,06:34:0 100.225 06/07/2016,10:02:0 101.023 100.9767 0.634222 1.638979 390 16.33936 9 0.400092

06/07/2016,13:04:0 101.457 07/07/2016,01:53:0 100.695 100.6961 -0.60257 -1.59805 12 0.229072

08/07/2016,08:30:0 100.241 08/07/2016,08:31:0 101.184 100.9928 0.252417 1.25431 1 4876.759 3 0.382866

08/07/2016,08:31:0 101.184 08/07/2016,09:10:0 100.395 100.4251 -0.52061 -1.51671 44 134.0219 1 0.010622

*******************************

In the 4th row, T = 1 which is correct [8:30 -> 8:31] - same day

but in the 5th row T = 44 which is incorrect, T should be 39 [8:31 -> 9:10] ??????

Most of the T values are completely wrong................

The whole exercise of calculating time differences is wrong due to the real life situation in the forex market

for example:- One price can be at 14:59 on a certain day whereas the very next price can be at 03:00 on the following day

Here the value of T is 1 ! because the prices are next to each other - forget subtracting dates and time

IMHO, the correct way is to look at the array of prices - which is the initial and only input to this exercise

By subtracting the index of the T_DCC Price from the index of the T_EXT Price - you get the correct 'time difference'

Bob M

[Walter Roberson]

In the extract you posted, as outside observers we do not have enough information to know what order the date/times were extacted from the file and stored into the Time array, and we also do not know what DCA is about.

[Bob Matthews]

Here is a snippet of the minute data input ( in chronological order)

****************************************

01/07/2016 00:00:00 102.827

01/07/2016 00:01:00 102.842

01/07/2016 00:02:00 102.843

01/07/2016 00:03:00 102.86

01/07/2016 00:04:00 102.865

01/07/2016 00:05:00 102.871

01/07/2016 00:06:00 102.881

01/07/2016 00:07:00 102.882

01/07/2016 00:08:00 102.875

01/07/2016 00:09:00 102.876

01/07/2016 00:10:00 102.869

*******************************************

I think DCA is the number of DC events (Directional Change Events)

where DC Events are Upturn, Downturn, Upturn, Downturn etc. (at a certain threshold)

Bob M