wsyn (version 1.0.4)

cleandat: Clean (spatio)temporal data matrices to make them ready for analyses using the wsyn package

Description

A data cleaning function for optimal Box-Cox transformation, detrending, standarizing variance, de-meaning

Usage

cleandat(dat, times, clev, lambdas = seq(-10, 10, by = 0.01), mints = NA)

Arguments

dat

A locations x time data matrix, or a time series vector (for 1 location)

times

The times of measurement, spacing 1

clev

The level of cleaning to do, 1 through 5. See details.

lambdas

A vector of lambdas to test for optimal Box-Cox transformation, if Box-Cox is performed. Ignored for clev<4. Defaults to seq(-10,10, by=0.01). See details.

mints

If clev is 4 or 5, then time series are shifted to have this minimum value before Box-Cox transformation. Default NA means use the smallest difference between consecutive, distinct sorted values. NaN means perform no shift.

Value

cleandat returns a list containing the cleaned data, clev, and the optimal lambdas from the Box-Cox procedure (NA for clev<4, see details).

Details

NAs, Infs, etc. in dat trigger an error. If clev==1, time series are (individually) de-meaned. If clev==2, time series are (individually) linearly detrended and de-meaned. If clev==3, time series are (individually) linearly detrended and de-meaned, and variances are standardized to 1. If clev==4, an optimal Box-Cox normalization procedure is applied jointly to all time series (so the same Box-Cox transformation is applied to all time series after they are individually shifted depending on the value of mints). Transformed time series are then individually linearly detrended, de-meaned, and variances are standardized to 1. If clev==5, an optimal Box-Cox normalization procedure is applied to each time series individually (again after individually shifting according to mints), and transformed time series are then individually linearly detrended, de-meaned, and variances are standardized to 1. Constant time series and perfect linear trends trigger an error for clev>=3. If clev>=4 and the optimal lambda for one or more time series is a boundary case or if there is more than one optimal lambda, it triggers a warning. A wider range of lambda should be considered in the former case.

References

Box, GEP and Cox, DR (1964) An analysis of transformations (with discussion). Journal of the Royal Statistical Society B, 26, 211<U+2013>252.

Venables, WN and Ripley, BD (2002) Modern Applied Statistics with S. Fourth edition. Springer.

Sheppard, LW, et al. (2016) Changes in large-scale climate alter spatial synchrony of aphid pests. Nature Climate Change. DOI: 10.1038/nclimate2881

See Also

wt, wmf, wpmf, coh, wlm, wlmtest, clust, browseVignettes("wsyn")

Examples

# NOT RUN {
times<-1:100
dat<-rnorm(100)
res1<-cleandat(dat,times,1) #this removes the mean
res2<-cleandat(dat,times,2) #detrends and removes the mean
res3<-cleandat(dat,times,3) #variances also standardized
res4<-cleandat(dat,times,4) #also joint Box-Cox applied
res5<-cleandat(dat,times,5) #1-3, also indiv Box-Cox

# }