Step 6: Where no man has gone before...

It was not enough for you, you want more WarpScript? You came at the right place, exoplanet hunter. This page is all about extracting a few WarpScript function that can be an help in your quest of exoplanets (or any other time series analysis)! You will discover functiond to establish a correlation between a series and a set of series, how to use the STL or the DISCORDS function, how to detect outliers in time series and some useful function to compute statistics on time series.

Those functions are listed here to give you a hint of the richness of WarpScript. At any time, feel free to skip this section to try to hack our code in the next step.

CORRELATE

The CORRELATE function can be used to compute the cross-correlation between a base time series and a list thereoff.

Here you have the example used in the documentation. An identical example as in the documtation is put below. You need a base time series, a list of the time series to compute the cross-correlation with and a third parameter corresponding to an offset list to compute the correlation. All time series must be correctly bucketized (aligned ticks) and filled (with no empty buckets) before, as the offset corresponds to a multiple of the distance between two points. The value 0 as offset will compute the correlation between all ticks of each series.

STL

The second function of this section is the well known STL (Seasonal and Trend decomposition using Losses). This function expects a list of two time series (or a list of list of two time series) onto the stack. These time series are the seasonal part and the trend part of the input time series. Seasonal and trend are added as suffix. You can look here for an introduction on how the STL decomposition works for time series. You will find the Warp10.io example below:

DISCORDS

The third section is about the DISCORDS functions that is based on pattern recognition to detect discontinuities in a time series of a list thereof. This functions expects some parameters to define how to recognize a pattern. The DISCORDS documentation explains more in details all the different parameters and present a working example.

A pattern is defined with first it’s window size, then the size of the pattern and finish with it’s quantization scale (values ladder dividing the window width). To compute the discord you need to add maximum number of discords to identify, a boolean flag indicating whether or not the DISCORDS function should report overlapping discords and a distance ratio (it is safe to leave it to 0.0). You will find the Warp10.io example below.

Outlier detection

In WarpScript, there is mutliple outlier function detection. You can retrieve for example the basic THRESHOLDTEST which just need a threshold value. This function will detect values greater than this threshold. There is also the HYBRIDTEST which mix STL and another detection method ESDTEST. You will find all the other available outlier method detection here. You will find below the Warp10.io example of the ESDTEST.

Some statisticals functions

WarpScript also have some basic statistics functions that can be apply on time series:

• MUSIGMA to compute both the mean and the standard deviation of a series
• NORMALIZE, ISONORMALIZE, STANDARDIZE
• SINGLEEXPONENTIALSMOOTHING to smooth and compare different series
• To extract more details information on a time series, you will also retrieve some functions as MODE or VALUEHISTOGRAM
• Many other statisticals functions are CPROB, LOWESS, LTTB, MONOTONIC, NSUMSUMSQ, PROB, RLOWESS and TLTTB, if needed one day.

Bonus, wanna a quick resume of some of the main WarpScript function?

Once you completed this tutorial, to remember some details of the WarpScript, feel free to use this Warp10 cheat sheet.

This is the end

Thanks for completed this tutorial! We were happy to introduce you to the analytics of time series worlds and to present a short initiation to our search of exoplanet!