olsTf: Ordinary least squares frequency-domain regression
In driegert/transfer: Frequency Domain Transfer Function Estimation Tools

Estimates coefficients by frequency (transfer function) between time-domain inputs x and response y.

1	olsTf(x, y, time, n, npredictor, ntaper, freq, fOffset)

`x`	is a `list` of `lists`: each sublist is a block of time containing numeric matrices like y - one matrix for each predictor
`y`	a `list` of complex matrices (containing real data probably) - each matrix is a block in time and contains response * dpss - n rows, k columns
`time`	A vector containing times at which the data are sampled for each block. (you can just supply 1:n)
`n`	An `integer` indicating the number of samples in each block.
`npredictor`	An `integer` indicating the number of inputs (X's)
`ntaper`	An `integer` indicating the number of tapers that were used.
`freq`	is a list of frequencies at which to estimate the transfer function - someone needs to take into account that there will be a gap at beginning and end of size max(fOffset).
`fOffset`	a vector indicating which frequencies of predictors to use (should at least be 0)

I need to revisit this and think of a way to NOT use the slow Fourier transform. Also, probably want to be using an SVD regression to get the transfer function.

driegert/transfer documentation built on May 15, 2019, 2:11 p.m.