How to create a new data set, from an existing data set, with a different time increment than the original

Suppose you have a very large x, y data set where "x" is time and "y" is some known value. The time increment for "x" is 0.02 seconds. "y" is known at every 0.02 seconds. I would like to change the time increment to 0.010269 seconds and use the known "y" values to find the new "y" values for the 0.010269 increment based off of a linear interpolation between the known original "y" values.

Answer 1

[jonas]

1. Convert x to duration array
2. Build a timetable with x and y
3. Use retime with timestep set to 0.010269

Provide data if you want code.

As a sidenote: based on the very specific timestep I'm guessing you want to align this time-series with another time-series. If that is the case, then simply use the duration array of the other time-series as input to retime.

[Austin Sowers]

Here is the data. The code will be very helpful. This sounds promising!

[jonas]

...and here is the code as promised

%%Load data
data = load('MATLAB Answers question data.mat');
x = data.dataset(:,1);
y = data.dataset(:,2);

%%Convert x to duration format and build timetable
TT1 = timetable(seconds(x),y);

%%Change timestep
TT2 = retime(TT1,'regular','linear','TimeStep',seconds(0.010269));

%%Display first rows
TT2(1:5,:)

ans =

5×1 timetable

       Time         y   
    __________    ______

    15 sec          4.31
    15.01 sec     4.2792
    15.021 sec    4.2511
    15.031 sec    4.2716
    15.041 sec    4.2873

It's a really ambiguous time-step, but whatever works for you!

[Austin Sowers]

Any idea what this error might be when I try using your code?

Error using timetable/retime (line 103) 'regular' is not a valid specification of target time vector for synchronization.

[Austin Sowers]

I have seen that it is because I am using version R2016b and not R2018b. I wonder what is the correct option to place for my version.

[jonas]

retime in R2016b only took character vectors, such as 'hourly', as new timestep. This method should however still work:

data = load('MATLAB Answers question data.mat');
x = data.dataset(:,1);
y = data.dataset(:,2);
t=seconds(x);
ts_new=seconds(0.010269);
t_new=t(1):ts_new:t(end);
TT1=timetable(t,y);
TT2 = retime(TT1,t_new','linear')

[Austin Sowers]

It looks like it worked! Thank you so much! I have been trying to do this for a week.

[jonas]

You're welcome! Duration (format), datetime (format) and timetables are extremely convenient when working with time-series.