Package 'zeitgebr'

Computes a spectrogram using CWT

Description

A port of Continuous Wavelet transform to rethomics. This function is intended to be used as an argument in the spectrogram wrapper.

Usage

cwt_spectrogram(x, period_range = c(hours(1), hours(32)),
  sampling_rate = 1/mins(1), resolution = 1/64,
  summary_time_window = mins(30))

Arguments

x	numeric vector
period_range	vector of size 2 defining minimal and maximal range of period to study (in seconds)
sampling_rate	the – implicitly regular – sampling rate of x (in hertz)
resolution	the period resolution of the CWT (i.e. the number of suboctaves)
summary_time_window	the sampling period after post-processing. Values of power are avegraged over this time window, for each period.

See Also

• spectrogram – to apply this fucntion to all indivvidual, with some preprocessing.

• WaveletComp::analyze.wavelet – the orginal function for cwt_spectrogram

---

A behavr table with approximately ten days of DAM2 recording for 32 fruit flies. The first 10, the following 11 and the last 11 animals have long, short and wild type period, respectively (see meta(dams_sample)).

Description

A behavr table with approximately ten days of DAM2 recording for 32 fruit flies. The first 10, the following 11 and the last 11 animals have long, short and wild type period, respectively (see meta(dams_sample)).

Usage

dams_sample

Format

An object of class behavr (inherits from data.table, data.frame) with 415040 rows and 3 columns.

Author(s)

Luis Garcia

References

Raw data stored at https://github.com/rethomics/zeitgebr/tree/master/raw_data

---

Find peaks in a periodogram

Description

This function locates the peaks in a pregenerated periodogram. Detection is based on pracma::findpeaks. Only the significant (i.e. power > signif_threshold) peaks are extracted.

Usage

find_peaks(data, n_peaks = 3)

Arguments

data	behavr::behavr table representing a periodogram, as returned by periodogram
n_peaks	maximal numbers of peak to be detected

Value

behavr::behavr table that is data with an extra column peak. peak is filled with zeros except for rows match a peak. In which case, rows have an integer value corresponding to the rank of the peak (e.g. 1 for the first peak).

References

• zeitgebr tutorial – the relevant rehtomics tutorial

See Also

• periodogram – to generate a periodogram in a first place

• ggetho::geom_peak – a layer to show peaks on a periodogram

Examples

data(dams_sample)
# only a half of the individuals for the sake of the example
dt <- dams_sample[xmv(region_id) %in% (1:16 * 2)]
per_dt_xs <- periodogram(activity, dt, FUN = chi_sq_periodogram)
per_dt_xs_with_peaks <- find_peaks(per_dt_xs)
per_dt_xs_with_peaks[peak == 1]

---

Computes periodograms

Description

This function builds periodograms, with one of several methods, for each individual of a behavr table

Usage

periodogram(var, data, period_range = c(hours(16), hours(32)),
  resample_rate = 1/mins(15), alpha = 0.01, FUN = chi_sq_periodogram,
  ...)

Arguments

var	variable to analyse
data	behavr table
period_range	vector of size 2 defining minimal and maximal range of period to study (in seconds)
resample_rate	frequency to resample (up or down) the data at (in hertz)
alpha	significance level
FUN	function used to compute periodogram (see periodogram_methods)
...	additional arguments to be passed to FUN

Value

A behavr::behavr table. In addition to the metadata, it contains data that encodes a periodogram (i.e. power vs period). The data contains the columns:

• power – the power the or equivalent (according to FUN)

• period – the period at which power is computed (in seconds)

• p_value – the p value associated to the power estimation

• signif threshold – the threshold above which power is considered significant

References

• zeitgebr tutorial – the relevant rehtomics tutorial

See Also

• periodogram_methods – the list of built-in methods

• find_peaks – to find peaks in the periodogram

• ggetho::ggperio – to plot periodograms

Examples

data(dams_sample)
# only a half of the individuals for the sake of the example
dt <- dams_sample[xmv(region_id) %in% (1:16 * 2)]
pdt <- periodogram(activity, dt, FUN = ls_periodogram, oversampling = 4)
pdt <- periodogram(activity, dt, FUN = chi_sq_periodogram)

require(ggetho)
ggperio(pdt, aes(colour=period_group)) + stat_pop_etho()

---

Methods For Computing Periodograms

Description

These functions provides a series of methods to assess periodicity of circadian processes.

Usage

ac_periodogram(x, period_range = c(hours(16), hours(32)),
  sampling_rate = 1/mins(1), alpha = 0.05)

chi_sq_periodogram(x, period_range = c(hours(16), hours(32)),
  sampling_rate = 1/mins(1), alpha = 0.05,
  time_resolution = hours(0.1))

cwt_periodogram(x, period_range = c(hours(16), hours(32)),
  sampling_rate = 1/mins(1), alpha = 0.05, resolution = 1/512,
  n_sim = 10)

fourier_periodogram(x, period_range = c(hours(16), hours(32)),
  sampling_rate = 1/mins(1), alpha = 0.05)

ls_periodogram(x, period_range = c(hours(16), hours(32)),
  sampling_rate = 1/mins(1), alpha = 0.05, oversampling = 8)

Arguments

x	numeric vector
period_range	vector of size 2 defining minimal and maximal range of period to study (in seconds)
sampling_rate	the – implicitly regular – sampling rate of x (in hertz)
alpha	significance level
time_resolution	the resolution of periods to scan
resolution	the period resolution of the CWT (i.e. the number of suboctaves)
n_sim	the number of shuffling simulation to compute p-value (see WaveletComp::analyze.wavelet)
oversampling	the oversampling factor (see lomb::lsp)

Value

a data.table with the columns:

• period – the period (in s)

• power – the power (or equivalent) for a given period

• p_value – the significance of the power

• signif_threshold – the significance threshold of the power (at alpha)

References

• zeitgebr tutorial – the relevant rehtomics tutorial

See Also

• lomb::lsp – the orginal function for ls_periodogram

• xsp::chiSqPeriodogram – code modified from

• stats::acf – the orginal function for ac_periodogram

• WaveletComp::analyze.wavelet – the orginal function for cwt_periodogram

---

Computes spectrogram

Description

This function builds spectrogram, using CWT, for each individual of a behavr table

Usage

spectrogram(var, data, period_range = c(hours(16), hours(32)),
  resample_rate = 1/mins(15), FUN = cwt_spectrogram, ...)

Arguments

var	variable to analyse
data	behavr table
period_range	vector of size 2 defining minimal and maximal range of period to study (in seconds)
resample_rate	frequency to resample (up or down) the data at (in hertz)
FUN	function used to compute spectrograms (so far, only CWT is implemented via cwt_spectrogram)
...	additional arguments to be passed to FUN

Details

A spectrogram is a estimation of the local periodicity of a signal at a given time. In the context of circadian rhythm, it can be useful to understand how infradian rhythms change along the day or, for instance, how circadian rhythm change ver the course of an multi-day experiment.

Value

A behavr::behavr table. In addition to the metadata, it contains data that encodes a spectrogram (i.e. power vs period). The data contains the columns:

• t – the time (in s) (same range the input time)

• period – the period at which the power is computed, for a given t (in s)

• power – the power the or equivalent (according to FUN)

• ridge – a logical defining whether the point (t and period) is a ridge

References

• spectrogram tutorial – the relevant rehtomics tutorial

See Also

• periodogram – to compute periodogram instead

• cwt_spectrogram – The dunction use to compute individual spectrograms

• ggetho::ggspectro – to plot spectrograms

Examples

data(dams_sample)
dt <- dams_sample[id %in% dams_sample[meta=TRUE, ,id[1:5]]]
spect_dt <- spectrogram(activity, dt)


require(ggetho)
ggspectro(spect_dt) +
        stat_tile_etho() +
        scale_y_log10() +
        facet_wrap(~ id)

---