reslr
Use:
# Not on CRAN yet
#install.packages("reslr")
#devtools::install_github("maeveupton/reslr")
install_github("maeveupton/reslr")
then,
library(reslr)
Note: The JAGS software is a requirement for this instruction sheet and refer back to main vignettes for more information.
reslr
There is a large example dataset included in the reslr
package called NAACproxydata
. In this example, we
demonstrate how to include proxy record data which is stored in a csv
file. This csv file of data can be found in the package and the
readr
function reads the csv file:
<- system.file("extdata", "one_data_site_ex.csv", package = "reslr")
path_to_data <- read.csv(path_to_data) example_one_datasite
Using the reslr_load
function to read in the data into
the reslr
package:
<- reslr_load(
example_one_site_input data = example_one_datasite)
plot(
x = example_one_site_input,
title = "Plot of the raw data",
xlab = "Year (CE)",
ylab = "Relative Sea Level (m)",
plot_tide_gauges = FALSE,
plot_caption = TRUE
)
Select your modelling technique from the modelling options available:
Statistical Model | Model Information | model_type code |
---|---|---|
Errors in variables simple linear regression | A straight line of best fit taking account of any age and measurement errors in the RSL values using the method of Cahill et al (2015). Use for single proxy site. | “eiv_slr_t” |
Errors in variables change point model | An extension of the linear regression modelling process. It uses piece-wise linear sections and estimates where/when trend changes occur in the data (Cahill et al. 2015). | “eiv_cp_t” |
Errors in variables integrated Gaussian Process | A non linear fit that utilities a Gaussian process prior on the rate of sea-level change that is then integrated (Cahill et al. 2015). | “eiv_igp_t” |
Noisy Input spline in time | A non-linear fit using regression splines using the method of Upton et al (2023). | “ni_spline_t” |
Noisy Input spline in space and time | A non-linear fit for a set of sites across a region using the method of Upton et al (2023). | “ni_spline_st” |
Noisy Input Generalised Additive model for the decomposition of the RSL signal | A non-linear fit for a set of sites across a region and provides a decomposition of the signal into regional, local-linear (commonly GIA) and local non-linear components. Again this full model is as described in Upton et al (2023). | “ni_gam_decomp” |
For this example, it is a single site and we are interested in how it varies over time select the Noisy Input spline in time. If it was multiple sites, we recommend using a spatial temporal model, i.e. Noisy Input spline in space and time, or for decomposing the signal, i.e. Noisy Input Generalised Additive model.
Once the model is chosen use the reslr_mcmc
function to
run it:
<- reslr_mcmc(
res_one_site_example input_data = example_one_site_input,
model_type = "ni_spline_t",
CI = 0.95
)
The convergence of the algorithm is examined and he parameter estimates from the model can be investigated using the following:
summary(res_one_site_example)
#> No convergence issues detected.
#> # A tibble: 2 × 7
#> variable mean sd mad q5 q95 rhat
#> <chr> <num> <num> <num> <num> <num> <num>
#> 1 sigma_beta 2.09 0.685 0.529 1.29 3.33 1.00
#> 2 sigma_y 0.00624 0.00489 0.00460 0.000447 0.0156 1.00
The model fit results can be visualised using the following function:
plot(res_one_site_example,
xlab = "Year (CE)",
ylab = "Relative Sea Level (m)",
plot_type = "model_fit_plot"
)
For the rate of change plot use:
plot(res_one_site_example,
plot_type = "rate_plot"
)
To examine the data creating these plots the user types the following:
<- res_one_site_example$output_dataframes
output_dataframes head(output_dataframes)
#> Longitude Latitude SiteName data_type_id Age pred
#> 1 -76.38 34.971 Cedar Island,\n North Carolina ProxyRecord -800 -2.312202
#> 2 -76.38 34.971 Cedar Island,\n North Carolina ProxyRecord -750 -2.316994
#> 3 -76.38 34.971 Cedar Island,\n North Carolina ProxyRecord -700 -2.317622
#> 4 -76.38 34.971 Cedar Island,\n North Carolina ProxyRecord -650 -2.314291
#> 5 -76.38 34.971 Cedar Island,\n North Carolina ProxyRecord -600 -2.307204
#> 6 -76.38 34.971 Cedar Island,\n North Carolina ProxyRecord -550 -2.296567
#> upr lwr rate_pred rate_upr rate_lwr CI
#> 1 -2.399767 -2.229546 -0.13882006 -0.65081467 0.3625245 95%
#> 2 -2.386640 -2.250235 -0.05351566 -0.48182156 0.3739311 95%
#> 3 -2.374637 -2.261549 0.02770941 -0.32980947 0.3860676 95%
#> 4 -2.364049 -2.262925 0.10485494 -0.19083120 0.4022410 95%
#> 5 -2.356266 -2.258588 0.17792093 -0.05568056 0.4203347 95%
#> 6 -2.345550 -2.247646 0.24690738 0.06597339 0.4350460 95%
To examine the additional options in the reslr
package,
see the main vignette.