com.yetanalytics.datasim.math.timeseries — com.yetanalytics/datasim 0.3.0

Timeseries namespaces; all timeseries are lazy, potentially infinite sequences of numeric values.

arma-seq clj

(arma-seq {:keys [std phi theta c seed] :as arma-model})

ARMA - AutoRegressive-Moving-Average - sequence generation.

An ARMA model describes a stochastic process in terms of two polynomials: the autogregression term and the moving average term. The model is written as such:

X_t = epsilon_t + SUM(phi_i X_t-i, i=1, p) + SUM(theta_i epsilon_t-i, i=1, q)

where X_t is the t-th value, epsilon_i are white noise parameters and phi_i and theta_i are the central parameters for the AR and MA models, respectively.

Besides :phi and :theta, which are colls of p and q double values respectively, the arma-model option map also has these additional params:

• :std, the standard deviation of the Gaussian distribution from which each epsilon_t is sampled from (the mean is fixed at zero)
• :seed, the seed to create the epsilon_t-generating RNG with.
• :c, a constant to add to each result X_t

Returns an infinite lazy seq of ARMA values.

hour-day-fraction clj

minute-day-fraction clj

ms-per-day clj

ms-per-hour clj

ms-per-minute clj

ms-per-second clj

ms-per-week clj

time-seqs clj

(time-seqs &
           {:keys [t-zero sample-ms zone]
            :or {t-zero 0 zone (t/zone-id "UTC")}})

Given a :t-zero (simulation start), an exclusive upper bound of :sample-ms milliseconds and an optional local time zone, return a map of useful lazy time sequences. (Note that since these are lazy seqs, we do not waste performance overhead on unused sequences.)

There are three kinds of sequences: sequences of milliseconds (intervals being equal to the time unit used); cyclic sequences of time with relation to the larger unit; and sinusoidal sequences of time with relation to the day-night cycle, where 1.0 represents midnight, -1.0 represents noon, positive numbers nighttime, and negative numbers daytime.

Time sequences in the returned map:

Sequence Key	Description
:milliseconds-seq	Sequence of epoch milliseconds since t-zero (e.g. (0 1 2 ...))
:second-ms-seq	Sequence of epoch ms with an interval of one second (e.g. (0 1000 ...))
:minute-ms-seq	Sequence of epoch ms with an interval of one minute (e.g. (0 60000 ...))
:hour-ms-seq	Sequence of epoch ms with an interval of one hour (e.g. (0 3600000 ...))
:day-ms-seq	Sequence of epoch ms with an interval of one day (e.g. (0 86400000 ...))
:week-ms-seq	Sequence of epoch ms with an interval of one week (e.g. (0 604800000 ...))
:minute-of-hour-seq	Cyclic sequence of minutes per hour (e.g. (0 1 ... 59 0 ...))
:minute-of-day-seq	Cyclic sequence of minutes per day (e.g. (0 1 ... 1440 0 ...))
:hour-of-day-seq	Cyclic sequence of hours per day (e.g. (0 1 ... 23 0 ...))
:day-of-week-seq	Cyclic sequence of the day of the week (e.g. (1 2 ... 7 1 ...))
:day-of-month-seq	Cyclic sequence of the day of the month (e.g. (1 2 ... 31 1 ...))
:day-of-year-seq	Cyclic sequence of the day of the year (e.g. (1 2 ... 365 1 ...))
:minute-day-night-seq	Sinusoidal sequence of milliseconds in relation to the day-night cycle.
:hour-day-night-seq	Sinusoidal sequence of milliseconds in relation to the day-night cycle.