Performing operations on non-uniform data

I want to perform subtraction operation on the second column of the below datapoints-

I want this operation to be performed on every 20 minutes. Basically, I am trying to find out log returns at every 20 minutes. However, since my data is not uniform, i am not able to write a general code to do the same. For example, i start my calculations at 00:00:00 and then go to 00:20:00 and my code works. However, when it goes to 01:00:00, it does not find any data. Also it does not find any data at 01:20:00 or 01:40:00.

%Original dataset
1512604800	'07-Dec-2017 00:00:00'	14090
1512604801	'07-Dec-2017 00:00:01'	14090
1512604803	'07-Dec-2017 00:00:03'	14090
1512604804	'07-Dec-2017 00:00:04'	14090
1512604805	'07-Dec-2017 00:00:05'	14090
1512604807	'07-Dec-2017 00:00:07'	14090
1512604808	'07-Dec-2017 00:00:08'	14090
1512604809	'07-Dec-2017 00:00:09'	14090
1512604810	'07-Dec-2017 00:00:10'	14090
1512604812	'07-Dec-2017 00:00:12'	14090
1512604813	'07-Dec-2017 00:00:13'	14090
1512604814	'07-Dec-2017 00:00:14'	14090
1512604815	'07-Dec-2017 00:00:15'	14090.5
1512604816	'07-Dec-2017 00:00:16'	14098
1512604817	'07-Dec-2017 00:00:17'	14098.98
1512604818	'07-Dec-2017 00:00:18'	14099
1512604820	'07-Dec-2017 00:00:20'	14098.99
1512604821	'07-Dec-2017 00:00:21'	14099
1512604822	'07-Dec-2017 00:00:22'	14098.99
1512604823	'07-Dec-2017 00:00:23'	14099
1512604825	'07-Dec-2017 00:00:25'	14099
1512604826	'07-Dec-2017 00:00:26'	14099

%Reduced dataset (Just for understanding the 20 min datapoints)
1512604800	'07-Dec-2017 00:00:00'	14090
1512606000	'07-Dec-2017 00:20:00'	14369.99
1512607206	'07-Dec-2017 00:40:06'	14389.99
1512608402	'07-Dec-2017 01:00:02'	14370.03
1512609601	'07-Dec-2017 01:20:01'	14388
1512610801	'07-Dec-2017 01:40:01'	14399.48
1512612000	'07-Dec-2017 02:00:00'	14361.01
1512613200	'07-Dec-2017 02:20:00'	13900
1512614400	'07-Dec-2017 02:40:00'	14079
1512615600	'07-Dec-2017 03:00:00'	13974.03
1512616804	'07-Dec-2017 03:20:04'	14082.94
1512618001	'07-Dec-2017 03:40:01'	13976
1512619201	'07-Dec-2017 04:00:01'	13911.8
1512620400	'07-Dec-2017 04:20:00'	14099.985


%I need these returns (have calcualted on excel sheet)
1512604800	'07-Dec-2017 00:00:00'	14090	        Desired_returns
1512606000	'07-Dec-2017 00:20:00'	14369.99	279.99
1512607206	'07-Dec-2017 00:40:06'	14389.99	20
1512608402	'07-Dec-2017 01:00:02'	14370.03	-19.96
1512609601	'07-Dec-2017 01:20:01'	14388	        17.97
1512610801	'07-Dec-2017 01:40:01'	14399.48	11.48
1512612000	'07-Dec-2017 02:00:00'	14361.01	-38.47
1512613200	'07-Dec-2017 02:20:00'	13900	        -461.01
1512614400	'07-Dec-2017 02:40:00'	14079	        179
1512615600	'07-Dec-2017 03:00:00'	13974.03	-104.97
1512616804	'07-Dec-2017 03:20:04'	14082.94	108.91
1512618001	'07-Dec-2017 03:40:01'	13976	        -106.94
1512619201	'07-Dec-2017 04:00:01'	13911.8	        -64.2
1512620400	'07-Dec-2017 04:20:00'	14099.985	188.185


%Code
startDate = final_data(1,1);
interval = 20; 
returnsForDay = []; 
previousValue = 14389.99; 
j=1; 
tempDateArray = []; 
flag = 0;
while j<=(length(final_data)) 
    startdatetemp = final_data(j,1); 
     CurrentDate = datestr( unixtime_to_datenum( startdatetemp ) ); 
     CurrentDateAfterInc = datestr( unixtime_to_datenum( startDate ) );  
     
    if(startDate == startdatetemp) 
        price_var = final_data(j,2);
        fprintf('Previous value when datesat are equal %d',previousValue); 
        if(j==1)
            previousValue = price_var;
        end
        newcol = length(returnsForDay);
        j
        returnsForDay(newcol + 1) = price_var - previousValue; 
        returnsForDay
        previousValue = price_var;
        flag = 1;
        j=j+1; 
    elseif(startDate > startdatetemp) 
        if(flag == 0)
            previousValue = final_data(j,2);
        end
        continue 
    elseif(startDate < startdatetemp) 
        newcol = length(returnsForDay); 
        fprintf('Previous value when datesat are not equal %d',previousValue); 
        if(flag ==1)
            fprintf('am here');
            returnsForDay(newcol + 1) = final_data(j,2) - previousValue;
        else
            if(j == 2)
                returnsForDay(newcol + 1) = previousValue - final_data(j-1,2); 
                returnsForDay
                previousValue = final_data(j-1,2); 
            else
                returnsForDay(newcol + 1) = previousValue - final_data(j-2,2); 
                returnsForDay
                previousValue = final_data(j-2,2);
            end %enf of small if statement
        end
    end %end of bigger if statement
   Converted_dates = datestr( unixtime_to_datenum( startDate ) ); 
   startdateTEMP = posixtime(datetime(Converted_dates) + minutes(00:interval:interval)); 
   startDate = startdateTEMP(1,2);  
end 

[dpb]

1. Attach a sample data file so folks have something can use to work with
2. Use timetable() and datetime not ancient datenum
3. retime() to minute keeping nearest value

Alternatively to 3. round (truncate, actually) to nearest minute first.

[Harsh Rob]

I have attached the dataset for one day for your reference.

Could you illustrate on the use of these suggestions?

Answer 1

[dpb]

t=timetable(datetime(final_data(:,1),'ConvertFrom','posixtime'),final_data(:,2));  % build timetable
t20=retime(t,'regular','next','TimeStep',minutes(20));                             % 20-minute w/ first after
t20(isnan(t20.Var1),:)=[];                                                         % rid of 08Dec 00:00:00 extrap

results in

>> [t20(1:10,:); t20(end-10:end,:)]
ans =
  21×1 timetable
            Time            Var1 
    ____________________    _____
    07-Dec-2017 00:00:00    14090
    07-Dec-2017 00:20:00    14370
    07-Dec-2017 00:40:00    14390
    07-Dec-2017 01:00:00    14370
    07-Dec-2017 01:20:00    14388
    07-Dec-2017 01:40:00    14399
    07-Dec-2017 02:00:00    14361
    07-Dec-2017 02:20:00    13900
    07-Dec-2017 02:40:00    14079
    07-Dec-2017 03:00:00    13974
    ....
    07-Dec-2017 20:20:00    15927
    07-Dec-2017 20:40:00    16130
    07-Dec-2017 21:00:00    16466
    07-Dec-2017 21:20:00    16778
    07-Dec-2017 21:40:00    16755
    07-Dec-2017 22:00:00    16750
    07-Dec-2017 22:20:00    16996
    07-Dec-2017 22:40:00    17213
    07-Dec-2017 23:00:00    17193
    07-Dec-2017 23:20:00    17098
    07-Dec-2017 23:40:00    17200
>> 

The difference is then simply

>> t20.Returns=[nan;diff(t20.Var1)]
t20 =
  72×2 timetable
            Time            Var1     Returns
    ____________________    _____    _______
    07-Dec-2017 00:00:00    14090        NaN
    07-Dec-2017 00:20:00    14370     279.99
    07-Dec-2017 00:40:00    14390         20
    07-Dec-2017 01:00:00    14370     -19.96
    07-Dec-2017 01:20:00    14388      17.97
    07-Dec-2017 01:40:00    14399      11.48
    07-Dec-2017 02:00:00    14361     -38.47
    07-Dec-2017 02:20:00    13900    -461.01
    07-Dec-2017 02:40:00    14079        179
    07-Dec-2017 03:00:00    13974    -104.97
    ...
    07-Dec-2017 20:00:00    15865     104.92
    07-Dec-2017 20:20:00    15927      61.65
    07-Dec-2017 20:40:00    16130      203.4
    07-Dec-2017 21:00:00    16466     336.35
    07-Dec-2017 21:20:00    16778     311.78
    07-Dec-2017 21:40:00    16755    -22.865
    07-Dec-2017 22:00:00    16750     -5.245
    07-Dec-2017 22:20:00    16996        246
    07-Dec-2017 22:40:00    17213     217.46
    07-Dec-2017 23:00:00    17193     -20.55
    07-Dec-2017 23:20:00    17098     -94.91
    07-Dec-2017 23:40:00    17200     101.98
>>