fitting-helpers: Trend-fitting helpers

In mkerschbaumer/mbte: Model-based trend estimation

Description

Helper functions meant to be used in conjunction with mbte_fit.

Usage

tr_linear()

tr_exponential()

tr_logistic()

Value

A fitting-quosure will be returned.

Functions

• tr_linear: Determine a linear trend using lm.

• tr_exponential: The presence of an exponenetial trend is determined by applying a log-transformation to the signal-values and fitting them via lm. A vector of the correct length with NA's will be returned if some values of the signal are zero or negative.

• tr_logistic: Fit a sigmoid model using nls2. This fitting helper requires the 'nls2'-package (the "port"-algorithm with 500 iterations is used). The signal-tibble to fit must have at least five rows (see mbte_fit). Otherwise NA's are returned.

  The formula used for fitting is value ~ A / (1 + exp(B * (C - time))) + D. NOTE: warnings are suppressed.

See Also

filtered_signals (dataset used in examples)

Examples

data(filtered_signals)

# Use fitting helpers to fit signals
#
# NOTE: The unquoting operator (`!!`) is used, since fitting helpers return
# quosures.
fits1 <- mbte_fit(filtered_signals, lin = !!tr_linear())
fits1

# Equivalent without using fitting-helpers (columns for time and value of
# signal have to be provided explicitly)
fits2 <- mbte_fit(filtered_signals, lin = lm(value ~ t, .signal))
fits2

mkerschbaumer/mbte documentation built on May 5, 2019, 11:01 p.m.