TASK LIST: Project 4 Ultradian analysis

Aggregate the interval data based on n-minute time intervals, return a time series. (standardize to per hour data)

def aggegate_interval(strain, mouse, feature, bin_width):
    """
    data loaded from data.load_intervals(feature)

    Parameters
    ---------------
    feature: {"AS", "Food", "IS", "M_AS", "M_IS", "Water", "Distance", "AS_Intensity", "AS_prob"}
    bin_width: number of minutes of time interval for data aggregation

    Returns
    ----------
    ts: pandas.tseries
        a pandas time series of length 12(day)*24(hour)*60(minute)/n
    """

Period detectiom: Lomb-Scargle Algorithm
- Reference Comparison of Lomb-Scargle implementation in Python Generalized vs Standard Lomb-Scargle in astroML package (with example of significance level calculation) Another link Example of Lomb-Scargle Algorithm in astroML package (with example of significance level calculation) Another link Lomb-Scargle in scipy package
- functions:

def plot_lomb_scargle(stain, mouse, day, feature, bin_width):
    """
    we detect ultradian period from 2 hours to 48 hours (to be decided)

    Returns
    ----------
    Periodograms: plot
        Lomb-Scargle Periodograms with horizontal line indicating 
        significance level = 0.05, 0.01, 0.001, etc.
    significant period: numpy.array
        period in number of bin width
    """

Seasonal decomposition
- Observed = Trend + Seasonal + Residuals
- fit time series model using residuals from the seasonal decomposition and detect the autocorrelation relationship
- Using residuals to calculate the SD/CI.
- Prediction of future features (F/W consumption) using Trend + Seasonal trem
- Reference Seasonal decomposition in statsmodels package, An implementation using statsmodel package
Mixed-effects model
- Reference: stats model implementation

berkeley-stat222 / mousestyles

TASK LIST: Project 4 Ultradian analysis #116