The venue for the meeting on Wednesday has changed - it is now in Room 145
on the 1st floor.
Speaker: Barry Quinn, UMIST
Title: Statistical Inference for Sinusoidal Regression Models
Time: 5.00pm (tea 4.30 pm) Wed 13th December 2000
Location: Room E145, First floor, John Dalton Building, Manchester
Metropolitan University
We all have some notion of `periodicity'. Some phenomena are more periodic
than others. A biological population may fluctuate with apparent
regularity, but this may as a result of a predator-prey relationship and
not associated with, say, an annual weather cycle. This behaviour
contrasts with, say, the noise generated by a motor because of constant,
or near-constant, rotation.
In many situations, the periodicity is unknown, and the problem of
interest it to estimate it accurately. This talk will mainly be concerned
with statistical inference for the frequency parameters in sinusoidal
regression models. It is natural to consider such models since any
periodic function can be arbitrarily approximated by sums of sinusoids
with frequencies that are multiples of a fundamental frequency. The talk
will be illustrated and partly motivated by problems arising in
(underwater passive sonar) signal processing. Because the analysis of such
data has to be done almost as soon as the data is received, techniques are
often needed that are sub-optimal with respect to statistical efficiency,
but orders of magnitude faster than optimal techniques.
|