var - Class: timeseries

Variance of timeseries data

Syntax

Description

ts_var = var(ts) returns the variance of ts.data.

ts_var = var(ts,Name,Value)uses additional options specified by one or more Name,Value pair arguments.

Input Arguments

ts	The timeseries object for which you want the variance of the data.

Name-Value Pair Arguments

'MissingData'	A string specifying one of two possible values, remove or interpolate, indicating how to treat missing data during the calculation. Default: remove
'Quality'	A vector of integers, indicating which quality codes represent missing samples (for vector data) or missing observations (for data arrays with two or more dimensions).
'Weighting'	A string specifying one of two possible values, none or time. When you specify time, larger time values correspond to larger weights.

Output Arguments

ts_var

The variance of ts.data, as follows: When ts.Data is a vector, then ts_var is the variance of ts.Data values. When ts.Data is a matrix, and IsTimeFirst is true, and the first dimension of ts is aligned with time, then ts_var is a row vector containing the variance of each column of ts.Data. When ts.Data is an N-dimensional array, var always operates along the first nonsingleton dimension of ts.Data.

Examples

The following example calculates the variance values of a multivariate timeseries object. MATLAB calculates the variance independently for each data column in the timeseries object.

% Load a 24-by-3 data array:

load count.dat

% Create a timeseries object with 24 time values:

count_ts = timeseries(count,[1:24],'Name','CountPerSecond');

% Calculate the variance of each data column for this timeseries object:
var(count_ts)

MATLAB returns:

1.0e+003 *
  0.6437    1.7144    4.6278

Algorithms

MATLAB determines weighting by:

1. Attaching a weighting to each time value, depending on its order, as follows:

   • First time point — The duration of the first time interval (t(2) - t(1)).

   • Time point that is neither the first nor last time point — The duration between the midpoint of the previous time interval to the midpoint of the subsequent time interval ((t(k + 1) - t(k))/2 + (t(k) - t(k - 1))/2).

   • Last time point — The duration of the last time interval ((t(end) - t(end - 1)).

2. Normalizing the weighting for each time by dividing each weighting by the mean of all weightings.

   Note   If the timeseries object is uniformly sampled, then the normalized weighting for each time is 1.0. Therefore, time weighting has no effect.

3. Multiplying the data for each time by its normalized weighting.

See Also

iqr | max | mean | median | min | std | sum | timeseries

© 1984-2012- The MathWorks, Inc.    -   Site Help   -   Patents   -   Trademarks   -   Privacy Policy   -   Preventing Piracy   -   RSS