Graph high frequency Forex data


Last week we learned how to use the date keyword to load dates into GAUSS. Today, we will plot some high-frequency Forex data.

The data

Today's dataset (usdcad_tick.csv) contains tick data for a little over 30,000 observations of the bid price for the USD/CAD currency pair from January 2, 2018.

This file has two variables:

  • timestamp: The time of the bid as a POSIX date/time (seconds since Jan 1, 1970).
  • bid: The bid price.

The first few observations of the file look like this:

1514872801.247, 1.25372
1514872801.497, 1.25376
1514872801.747, 1.25373
1514872802.247, 1.25377
1514872802.497, 1.25377

The fractional portion of the timestamp data represents nanoseconds.


plotTSHF is a function which will draw graphs of irregularly spaced and high-frequency time series data. It has three required inputs:

Tx1 vector containing POSIX date/times for each observation.
The unit to use for X-axis tick labels. Valid options include:
  • milliseconds
  • seconds
  • minutes
  • hours
  • days
  • months
  • quarters
  • years
Tx1 vector or TxP matrix containing the data to plot.

Basic plot

We can plot our dataset with the following code:

// Load all observations of both variables
fname = "usdcad_tick.csv";
data = loadd(fname);

// Split variables into separate vectors
timevec = data[.,1];
bid = data[.,2];

// Plot data
plotTSHF(timevec, "hours", bid);

and we will see a graph that looks like this:

Graph of high-frequency USD/CAD currency pairs.


Most of the code we used to create this graph should be fairly straightforward, with the possible exception of the label_unit input.

plotTSHF(timevec, "hours", bid);

In the above line of code, the label_unit is "hours". This tells GAUSS that we would like X-tick labels to be set at even hours.

If we used "days" we would not see any X-tick labels. This is because our data is from 06:00-18:00 on the same day. Since the time series data does not cross from one day to another, the graph does not contain the start of a day at which to place the tick label.

Customize tick labels

Now we will use plotSetXTicInterval to set the X-ticks to appear every 60 minutes, starting at 06:30.

To accomplish this, we will have to take the following steps:

  1. Set the label_unit passed to plotTSHF to "minutes".
    • This step tells plotTSHF to place the X-tick labels at even minutes.
  2. Set the tick_interval input to plotSetXTicInterval to be 60.
    • Since the tick_interval is in terms of the plotTSHF label_unit, setting tick_interval to 60 will set the X-ticks to be placed every 60 minutes.
  3. Set the first_labeled input to plotSetXTicInterval to be a POSIX date which equals Jan 2, 2018 06:30.
    • Since plotTSHF uses only POSIX date/times, we must specify the first X-tick to label as a POSIX date/time.
// Load all observations of both variables
fname = "usdcad_tick.csv";
data = loadd(fname);

// Split variables into separate vectors
timevec = data[.,1];
bid = data[.,2];

// Declare 'myPlot' to be a plotControl struct
// and fill with default values
struct plotControl myPlot;
myPlot = plotGetDefaults("xy");

// Convert the DT Scalar date to POSIX
// 2018 01 02 06 30 00
first_lbl = dttoposix(2018010206300);

// Draw X-tick labels every 60 minutes,
// starting at Jan 2, 2018 at 06:30
plotSetXTicInterval(&myPlot, 60, first_lbl);

// Plot data (notice we pass in the plotControl struct)
plotTSHF(myPlot, timevec, "minutes", bid);

The above code will create a graph like shown below.

Graph of high-frequency USD/CAD currency pair data.


Congratulations! You have learned how to:

  1. Plot high-frequency time series data with plotTSHF, using POSIX date/times.
  2. Customize the frequency and starting point of X-tick labels for time series graphs using plotTSHF.

Code and data from this blog can be found here.

Leave a Reply