wq (version 0.4.8)

mannKen: Mann-Kendall trend test and the Sen slope

Description

Applies Kendall's tau test for the significance of a monotonic time series trend (Mann 1945). Also calculates the Sen slope as an estimate of this trend.

Usage

mannKen(x, plot = FALSE, type = c("slope", "relative"), order = FALSE, pval = .05, pchs = c(19, 21), ...)

Arguments

x

A numeric vector, matrix or data frame

plot

Should the trends be plotted when x is a matrix or data frame?

type

Type of trend to be plotted, actual or relative

order

Should the plotted trends be ordered by size?

pval

p-value for significance

pchs

Plot symbols for significant and not significant trend estimates, respectively

...

Other arguments to pass to plotting function

Value

A list of the following if x is a vector:

sen.slope

Sen slope

sen.slope.rel

Relative Sen slope

p.value

Significance of slope

S

Kendall's S

varS

Variance of S

miss

Fraction of missing slopes connecting first and last fifths of x

or a matrix with corresponding columns if x is a matrix or data frame.

Details

The Sen slope (alternately, Theil or Theil-Sen slope)---the median slope joining all pairs of observations---is expressed by quantity per unit time. The fraction of missing slopes involving the first and last fifths of the data are provided so that the appropriateness of the slope estimate can be assessed and results flagged. Schertz et al. [1991] discuss this and related decisions about missing data. Other results are used for further analysis by other functions. Serial correlation is ignored, so the interval between points should be long enough to avoid strong serial correlation.

For the relative slope, the slope joining each pair of observations is divided by the first of the pair before the overall median is taken. The relative slope makes sense only as long as the measurement scale is non-negative (not, e.g., temperature on the Celsius scale). Comparing relative slopes is useful when the variables in x have different units.

If plot = TRUE, then either the Sen slope (type = "slope") or the relative Sen slope (type = "relative") are plotted. The plot symbols indicate, respectively, that the trend is significant or not significant. The plot can be customized by passing any arguments used by dotchart such as xlab or xlim, as well as graphical parameters described in par.

References

Mann, H.B. (1945) Nonparametric tests against trend. Econometrica 13, 245--259.

Helsel, D.R. and Hirsch, R.M. (2002) Statistical methods in water resources. Techniques of Water Resources Investigations, Book 4, chapter A3. U.S. Geological Survey. 522 pages. http://pubs.usgs.gov/twri/twri4a3/

Schertz, T.L., Alexander, R.B., and Ohe, D.J. (1991) The computer program EStimate TREND (ESTREND), a system for the detection of trends in water-quality data. Water-Resources Investigations Report 91-4040, U.S. Geological Survey.

See Also

seaKen, seasonTrend, tsSub

Examples

tsp(Nile)  # an annual time series
mannKen(Nile)

y <- sfbayChla
y1 <- interpTs(y, gap=1)  # interpolate single-month gaps only
y2 <- aggregate(y1, 1, mean, na.rm=FALSE)
mannKen(y2)
mannKen(y2, plot=TRUE) # missing data means missing trend estimates
mannKen(y2, plot=TRUE, xlim = c(0.1, 0.25))
mannKen(y2, plot=TRUE, type='relative', order = TRUE, pval = .001,
  xlab = "Relative trend")
legend("topleft", legend = "p < 0.001", pch = 19, bty="n")