sars (version 2.0.0)

threshold_ci: Calculate confidence intervals around breakpoints

Description

Generate confidence intervals around the breakpoints of the one-threshold continuous and left-horizontal models. Two types of confidence interval can be implemented: a confidence interval derived from an inverted F test and an empirical bootstrap confidence interval.

Usage

threshold_ci(object, cl = 0.95, method = "boot", interval = NULL,
  Nboot = 100, verb = TRUE)

Value

A list of class "sars" with two elements. If method “F” is used, the list contains only the confidence interval values. If method “boot” is used, the list contains two elements. The first element is the full set of bootstrapped breakpoint estimates for each model and the second contains the confidence interval values.

Arguments

object

An object of class 'thresholds', generated using the sar_threshold function. The object must contain fits of either (or both) of the one-threshold continuous or the one-threshold left-horizontal model.

cl

The confidence level. Default value is 0.95 (95 percent).

method

Either bootstraping (boot) or inverted F test (F).

interval

The amount to increment the threshold value by in the iterative model fitting process used in both the F and boot methods. The default for non-transformed area reverts to 1, while for log-transformed area it is 0.01. It is advised that the same interval value used when running sar_threshold is used here.

Nboot

Number of bootstrap samples (for use with method = "boot").

verb

Should progress be reported. If TRUE, every 50th bootstrap sample is reported (for use with method = "boot").

Author

Francois Rigal and Christian Paroissin

Details

Full details of the two approaches can be found in Toms and Lesperance (2003). If the number of bootstrap samples is large, the function can take a while to run. Following Toms and Lesperance (2003), we therefore recommend the use of the inverted F test confidence interval when sample size is large, and bootstrapped confidence intervals when sample size is smaller.

Currently only available for the one-threshold continuous and left- horizontal threshold models.

References

Toms, J.D. & Lesperance, M.L. (2003) Piecewise regression: a tool for identifying ecological thresholds. Ecology, 84, 2034-2041.

Examples

data(aegean2)
a2 <- aegean2[1:168,]
fitT <- sar_threshold(data = a2, mod = "ContOne", 
interval = 0.1, non_th_models = TRUE, logAxes = "area", logT = log10)
#calculate confidence intervals using bootstrapping
#(very low Nboot just as an example)
CI <- threshold_ci(fitT, method = "boot", interval = NULL, Nboot = 3)
CI
#Use the F method instead, with 90% confidence interval
CI2 <- threshold_ci(fitT, cl = 0.90, method = "F", interval = NULL)
CI2