Data, due to McNeall, from 92 runs of a climate model
McNeall used a numerical climate model and ran it 92 times, on a design matrix specified on 16 independent variables as detailed in McNeall 2008.
The model output is a temperature distribution over the surface of the
Earth. The model gives 2048 temperatures, corresponding to 2048 grid
squares distributed over the Earth. A vector of 2048 temperatures may
be displayed on a global map using the
The 92 model runs are presented in the form of a 2048 by 92 matrix
mcneall_temps, each column of which corresponds to a run. A row
of 92 temperatures corresponds to the temperature at a particular place
on the earth as predicted by each of the 92 model runs.
Following McNeall, a principal component analysis on the maps was
performed. The first four were used. Matrix
eigenmaps is a 2048
by 4 matrix, with columns corresponding to the four principal
mcneall_pc is a 92-by-20 matrix. The first 16 columns
correspond to the independent variables (ie the design matrix); columns
17-20 correspond to the first four principal components of the model
output. The 92 rows correspond to the 92 model runs.
The package can be used on the
mcneall_temps matrix; use
apart() to generate a
mdm object. A reasonably optimized
hyperparameters object of class
mhp is given as
D. McNeall 2008. "Dimension Reduction in the Bayesian analysis of a numerical climate model". PhD thesis, University of Southampton.
1 2 3