Package for spatio-temporal modelling. Contains functions that estimate, simulate and predict from the model described in (Szpiro et.al., 2010; Sampson et.al., 2011; Lindstrom et.al., 2010). The package also contains functions that handle missing data SVD in accordance with (Fuentes et.al. 2006).
| Package: | SpatioTemporal |
| Type: | Package |
| Version: | 1.1.9 |
| Date: | 2018-06-20 |
| License: | GPL version 2 or newer |
| LazyLoad: | yes |
Examples in the package uses data from the Multi-Ethnic Study of Atherosclerosis and Air Pollution (MESA Air), (Cohen et.al.,2009).
Upates: R 3.5.0 Compatibility and Matrix
Minor updates to fullfill R 3.5.0 and changes to Matrix-package.
Upates: R 3.2.1 Compatibility
Minor updates to fullfill R 3.2.1 changes.
Upates: Handling of log-Gaussian fields
Updated several functions to allow for prediction and CV of
log-Gaussian fields. Updated functions:
predict.STmodel, print.predictSTmodel,
plot.predictSTmodel,
predictCV.STmodel, print.predCVSTmodel,
summary.predCVSTmodel, plot.predCVSTmodel,
qqnorm.predCVSTmodel, and
scatterPlot.predCVSTmodel.
Updated predict.STmodel to compute temporal
averages, and return both prediction and variance of the
averages. Both for Gaussian and log-Gaussian data.
Upates: sparse-Matrices and temporal basis functions
Allows for sparse matrices in makeSigmaB and
makeSigmaNu; this reduces the memory footprint and
execution time for loglikeST,
predict.STmodel, and estimate.STmodel.
Added function that does regression estimates of the
beta-coefficients: estimateBetaFields.
Altered computation of CV-statistics in SVDsmoothCV.
Added boxplot.SVDcv for illustration of CV-statistics
from SVDsmoothCV.
Replaced updateSTdataTrend with
updateTrend.STdata and
updateTrend.STmodel that also allows for temporal
trends defined using functions.
Updated SVDsmooth, SVDsmoothCV, and
calcSmoothTrends to return both the trend and the
smoothing function used to compute the trends, simplifying
interpolation at unobserved time-points.
Updated example data-sets.
Added options for computation of temporal averages
(incl. variances) to predict.STmodel and
predictCV.STmodel.
Major bug fixes:
In predict.STmodel, predictions now always
uses the trend given in object, ignoring the trend object
in STdata. Prediction at dates in STdata are
computed using the smoothing function that defines the trend; see
updateTrend.STmodel for details.
In summary.predCVSTmodel, code previously divided by
the wrong variance when computing adjusted R2 using the
pred.naive option.
In summary.predCVSTmodel, code previously
returned statistics even for dates without observations when
using by.date=TRUE.
In plot.STdata and plot.STmodel code
now accounts for missing time-points when computing acf and pacf.
Added plot funcions/Minor fixes:
Added scatterPlot.STdata,
scatterPlot.STmodel,
and scatterPlot.predCVSTmodel for plotting
observations/residuals against covariates.
Added plot.mcmcSTmodel,
density.mcmcSTmodel, and
plot.density.mcmcSTmodel for plotting of MCMC
results.
Added qqnorm.STdata, qqnorm.STmodel,
and qqnorm.predCVSTmodel for plotting of data and
CV-prediction results.
Added a restart option to estimate.STmodel
allowing for restarts of optimisation in cases on bad
optimisation.
Minor changes/Bug fixes:
Fixed stupid misstake in predictNaive that caused
computations to take unnecessarily long.
Minor changes/Bug fixes:
Fixed a bug in SVDsmooth, that caused the values in
the temporal smooths to depend on the number of unobserved
time points.. This also affects
calcSmoothTrends and updateSTdataTrend
when the option extra.dates is in use.
Fixed bug in simulate.STmodel that caused NA
values when simulating at unobserved sites.
Fixed bug in predict.STmodel that could cause
errors when predicting at unobserved sites.
Fixed bug in predictCV.STmodel and
predict.STmodel; these will now handle predictions
at locations with incomplete nugget covariates.
Updated c.STmodel and predict.STmodel
to avoid errors/warnings due to more complex nugget models.
Replaced warning in createSTdata when
extra.dates!=NULL and n.basis=NULL with a message.
Bug fixes:
c.STmodel will now combine STmodel
objects with identical covariate scaling.
Major Changes:
Changed the return of the variances for beta in
predict.STmodel.
Reduced the memory footprint of predict.STmodel.
Error checks in c.STmodel and
predict.STmodel, combination of STmodel
objects with different covariate scaling is NOT
possible.
Added:
New plot function: plot.predCVSTmodel.
coef.estimateSTmodel and
coef.estCVSTmodel functions that extract estimated
parameters.
Parameters for predict.STmodel and
predictCV.STmodel can be specified using
estimateSTmodel or estCVSTmodel objects.
An lwd option to plot.predictSTmodel.
A short introductory vignette as complement to the full tutorial.
Bug fixes:
predictNaive now works for only one locations.
detrendSTdata now works for different regions.
Added packages maps and plotrix to suggested packages.
Bug fixes:
prediction for leave-one-out CV. stop updateCovf crashing in Rscript/R CMD BATCH.
Minor bug fixes
Updated documentation and vignette
Major change, most old functions are now deprecated. New features:
Different covariance functions Nuggets in the beta-fields Different nuggets for different locations in the nu-field. Different coordinates for beta and nu-fields, allowing for precomputed deformations Covariates can be specifed using formula-objects
Minor updates - no user visible changes
First CRAN-release
First released version, short course at TIES-2010
M. A. Cohen, S. D. Adar, R. W. Allen, E. Avol, C. L. Curl, T. Gould, D. Hardie, A. Ho, P. Kinney, T. V. Larson, P. D. Sampson, L. Sheppard, K. D. Stukovsky, S. S. Swan, L. S. Liu, J. D. Kaufman. (2009) Approach to Estimating Participant Pollutant Exposures in the Multi-Ethnic Study of Atherosclerosis and Air Pollution (MESA Air). Environmental Science & Technology: 43(13), 4687-4693.
M. Fuentes, P. Guttorp, and P. D. Sampson. (2006) Using Transforms to Analyze Space-Time Processes in Statistical methods for spatio-temporal systems (B. Finkenstadt, L. Held, V. Isham eds.) 77-150
J. Lindstrom, A. Szpiro, P. D. Sampson, L. Sheppard, A. Oron, M. Richards, and T. Larson T. (2010) A flexible spatio-temmporal model for air pollution: allowing for spatio-temporal covariates. Berkeley Electronic Press, University of Washington Biostatistics Working Paper Series, No. 370. http://www.bepress.com/uwbiostat/paper370
A. Szpiro, P. D. Sampson, L. Sheppard, T. Lumley, S. D. Adar, and J. D. Kaufman. (2010) Predicting intra-urban variation in air pollution concentrations with complex spatio-temporal dependencies. Environmetrics: 21, 606-631.
P. D. Sampson, A. Szpiro, L. Sheppard, J. Lindstrom, J. D. Kaufman. (2011) Pragmatic Estimation of a Spatio-temporal Air Quality Model with Irregular Monitoring Data. Atmospheric Environment: 45(36), 6593-6606.
# NOT RUN {
##For a short introduction see:
# }
# NOT RUN {
vignette("ST_intro",package="SpatioTemporal")
# }
# NOT RUN {
##For a worked out data-analysis exmaple see the tutorial.
##NOTE: This vignette is still work in progress
# }
# NOT RUN {
vignette("Tutorial",package="SpatioTemporal")
# }
# NOT RUN {
# }
Run the code above in your browser using DataLab