Statistics Seminar - University of Edinburgh

[Colin Aitken-2]

Statistics Seminar

School of Mathematics
The University of Edinburgh

Monday 18th August

2.00 p.m. Room 4312, James Clerk Maxwell Building

Tim Swartz,
Department of Statistics and Actuarial Science
Simon Fraser University


Tweaking Some Classical Methodologies into Bayesian Versions
               

With the advent of personal computing and the development of simulation
algorithms, there has been an explosion of Bayesian work carried out
over the last 20 years. Although much of the work concerns new
methodologies, many Bayesian developments have been the outgrowth of
longstanding classical procedures. In this talk, I describe two projects
in this direction where in each case the Bayesian benefits are
highlighted. The first project concerns a Bayesian implementation of the
agglomerative clustering algorithm also known as the unweighted paired
group method of averaging (UPGMA). The Bayesian version is applied to
datasets in adjudication and in stylometry. The second project concerns
the development of various longstanding social network models. In both
problems, MCMC methods are utilized.


Tea and coffee will be available after the seminar in the Mathematics
Common Room (5212).


Any enquiries about this seminar should be made to

Martin Delaney, James Clerk Maxwell Building, Room 6314.
Phone: [phone]
E-mail: martin.delaney@....



--
Professor Colin Aitken,
Professor of Forensic Statistics,
School of Mathematics, King’s Buildings, University of Edinburgh,
Mayfield Road, Edinburgh, EH9 3JZ.

Tel:    [phone]
E-mail:  c.g.g.aitken@...
Fax :  [phone]
http://www.maths.ed.ac.uk/~cgga


The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

[Natalia Bochkina]

Statistics Seminar

School of Mathematics
The University of Edinburgh

Friday 10 October 2008

3.00 p.m. Room 5327, James Clerk Maxwell Building


Chris Glasbey (Biomathematics and Statistics Scotland)

Spatio-temporal weather models
(joint with Dave Allcroft)


We develop contrasting spatio-temporal models for two weather variables:
solar radiation and rainfall.  For solar radiation the aim is to assess
the performance of area networks of photo-voltaic cells.  Although
radiation measured at a sufficiently fine temporal scale has a bimodal
marginal distribution (Glasbey, 2001), averages of 10-minute or longer
duration can be transformed to be approximately Gaussian, and we fit a
spatio-temporal auto-regressive moving average (STARMA) process (Glasbey
and Allcroft, 2008).  For rainfall, the aim is to disaggregate to a
finer spatial scale than that observed.  To overcome the difficulty that
the marginal distribution of hourly rainfall has a singularity at zero
and so is highly non-Gaussian, we apply a monotonic transformation.
This defines a latent Gaussian variable, with zero rainfall
corresponding to censored values below a threshold, which we model using
a spatio-temporal Gaussian Markov random field (Allcroft and Glasbey,
2003).  For both models, computations are simplified by approximating
space by a torus and using Fourier transforms.

Allcroft, D.J.  and Glasbey, C.A.  (2003).  A latent Gaussian Markov
random field model for spatio-temporal rainfall disaggregation.  Applied
Statistics, 52, 487-498.

Glasbey CA (2001). Nonlinear autoregressive time series with
multivariate Gaussian mixtures as marginal distributions.  Applied
Statistics, 50, 143-154.

Glasbey, C.A.  and Allcroft, D.J.  (2008).  A STARMA model for solar
radiation.  Applied Statistics, 57, 343-355.


Tea and coffee will be available after the seminar in the Mathematics  
Common Room (5212).

Natalia Bochkina
N.Bochkina@...


--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.

[John Bibby]

Sorry I'm unable to get to Edinburgh for this(but must find an excuse soon)

It prompted me to ask the weather question we have all asked but I have
never had a definitive answer: When we hear "Probability of precipitation in
x on Sunday is y%", what does this mean?

Is this statement a sum over all x (e.g. all UK) and all of Sunday i.e. the
event that it will precipitate somewhere in UK sometime on Sunday? I can't
see anything else that make sense, but I wd expect such probabilities to be
close to 100% on virtually every day (and they are not).

On the web I have found contradictory responses: e.g. sum over x, sum over
Sunday, and sum over both.

I'd appreciate any answer with web-link that will put my anxious mind at
rest! :)
 
     Best Regards
     JOHN BIBBY    
           for aa42.com Limited - International Data, Education and Trade
Consultants
           Please visit www.aa42.com/mathemagic
                   1 Straylands Grove, York YO31 1EB  (01904-330-334)
-----Original Message-----
From: A UK-based worldwide e-mail broadcast system mailing list
[mailto:allstat@...] On Behalf Of Natalia Bochkina
Sent: 08 October 2008 13:43
To: allstat@...
Subject: Statistics Seminar - University of Edinburgh

Statistics Seminar

School of Mathematics
The University of Edinburgh

Friday 10 October 2008

3.00 p.m. Room 5327, James Clerk Maxwell Building


Chris Glasbey (Biomathematics and Statistics Scotland)

Spatio-temporal weather models
(joint with Dave Allcroft)


We develop consting spatio-temporal models for two weather variables:
solar radiation and rainfall.  For solar radiation the aim is to assess
the performance of area networks of photo-voltaic cells.  Although
radiation measured at a sufficiently fine temporal scale has a bimodal
marginal distribution (Glasbey, 2001), averages of 10-minute or longer
duration can be transformed to be approximately Gaussian, and we fit a
spatio-temporal auto-regressive moving average (STARMA) process (Glasbey
and Allcroft, 2008).  For rainfall, the aim is to disaggregate to a
finer spatial scale than that observed.  To overcome the difficulty that
the marginal distribution of hourly rainfall has a singularity at zero
and so is highly non-Gaussian, we apply a monotonic transformation.
This defines a latent Gaussian variable, with zero rainfall
corresponding to censored values below a threshold, which we model using
a spatio-temporal Gaussian Markov random field (Allcroft and Glasbey,
2003).  For both models, computations are simplified by approximating
space by a torus and using Fourier transforms.

Allcroft, D.J.  and Glasbey, C.A.  (2003).  A latent Gaussian Markov
random field model for spatio-temporal rainfall disaggregation.  Applied
Statistics, 52, 487-498.

Glasbey CA (2001). Nonlinear autoregressive time series with
multivariate Gaussian mixtures as marginal distributions.  Applied
Statistics, 50, 143-154.

Glasbey, C.A.  and Allcroft, D.J.  (2008).  A STARMA model for solar
radiation.  Applied Statistics, 57, 343-355.


Tea and coffee will be available after the seminar in the Mathematics  
Common Room (5212).

Natalia Bochkina
N.Bochkina@...


--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.