Computes exponentially weighted moving average of a series.


y = movingaveExpwgt(x, d, p)
  • x (NxK matrix) – data
  • d (scalar) – order of moving average.
  • p (scalar) – smoothing coefficient where \(0 > p > 1\).

y (NxK matrix) – filtered series. The first \(d-1\) rows of x are set to missing values.


Smoothing of random walk data

// Create random data, setting initial seed
// value to be the hundredths of a second
// since midnight
numPoints = 1000;
seed = hsec;
{ delta, state } = rndn(numPoints, 1, seed);

// Set mean of data to be > 0
// to give data long term positive bias
delta = delta + 0.01;

// Increase magnitude of delta to
// create desired level of volatiliy
delta = 2*delta;

// Instantiate y : index price data
y = 1000*ones(numPoints,1);

// Loop through y and add the cumulative
// sums of delta to create random walk
for i(2, numPoints, 1);
  y[i]= y[i-1] + delta[i];

// Find moving average
twentyMA = movingaveExpwgt(y, 20, 0.8);


movingaveExpwgt() is smoothing time series filter using exponential weights. The moving average as performed by column and thus it treats the NxK matrix as K time series of length N.

See also

Functions movingaveWgt(), movingave()