amt (version 0.0.7)

adjust_param: Adjust parameters

Description

Functions for parameter adjustment after fitting an integrated step-selection function (iSSF).

Usage

adjust_shape(tentative, modifier = 0)
adjust_scale(tentative, modifier = 0)
adjust_kappa(tentative, modifier = 0)

Arguments

tentative

[numeric] The tentative parameter estimate.

modifier

[numeric=0] The modifier to adjust the tentative estimate.

Details

The shape and scale parameter of a gamma distribution, and the concentration parameter (=kappa) of a von-Mises distribution can be adjusted. The following adjustments are possible:

1. The shape parameter of a gamma distribution fitted to the observed step lengths, can be adjusted with the coefficient for the log of the step lenghts.

2. The scale parameter of a gamma distribution fitted to the observed step lengths, can be adjusted with the coefficient for the step lenghts.

3. The concentration parameter of fa von Mises distribution fitted to the observed turning angle, can be adjusted with the coefficient for the cosine of turning angles.

References

avgar2016amt

Examples

# NOT RUN {
# Using the deer data set
data(deer)
data(sh_forest)

# first prepare the data and fit a model
m1 <- deer %>% steps_by_burst() %>%
  random_steps() %>%
  extract_covariates(sh_forest) %>%
  mutate(sh.forest = factor(sh.forest)) %>%
  fit_clogit(case_ ~ sh.forest * log(sl_) + sl_ + strata(step_id_))



# Investigate and adjust parameters ---------------------------------------

sl_params(m1)['shape', 'est']
# adjust shape with the log of the step length
sh1 <- adjust_shape(sl_params(m1)['shape', 'est'],
             modifier = coef(m1)['log(sl_)'])

sl_params(m1)['scale', 'est']
# adjust shape with the step length
sc1 <- adjust_shape(sl_params(m1)['scale', 'est'],
             modifier = coef(m1)['sl_'])

# Up to now we have ignored the interaction with forest
# this means the above assumes that forest = 2 (= non forest)

sl_params(m1)['shape', 'est']
# adjust shape with the log of the step length
sh2 <- adjust_shape(tentative = sl_params(m1)['shape', 'est'],
                    modifier = coef(m1)['log(sl_)'] + coef(m1)['sh.forest2:log(sl_)'])

# The modified shape paremeter differ for forest and non forest.
# The shape for steps that end in forest are lower.
sh1
sh2

# This can be best seen when plotting the tentative Gamma distribution (black) and
# adding lines for Gamma distributions with adjusted shape paremeters
# for open areas (red) and forested areas (green).
# }
# NOT RUN {
plot_sl(m1)
curve(dgamma(x, shape = sh1, scale = sc1), col = "red", add = TRUE, from = 0.1)
curve(dgamma(x, shape = sh2, scale = sc1), col = "forestgreen", add = TRUE, from = 0.1)
# }
# NOT RUN {
# }