The goal of profrep is to calculate individual profile repeatability (Reed et al., 2019).
One of the most common measurements that stress physiologists take is blood samples for corticosterone quantification during a stress response. This typically includes a sample at baseline (<3 minutes of stressor onset), one or more stress-induced samples (e.g. 30 minutes after stressor onset), and potentially a negative feedback sample (e.g. 120 minutes after stressor onset and/or after dexamethasone injection). Such time series are called “stress response curves” and may be taken multiple times in one individual. If researchers have multiple stress response curves for an individual, they may want to quantify repeatability to investigate, for example, heritibility. The current standard in the field is to use linear mixed-effect models (Baugh et al. 2014; Dingemanse and Dochtermann, 2013), however this type of repeatability estimate can only be done on populations and on only one timepoint at a time. Reed et al. (2019) have proposed “Profile Repeatability,” which uses the full stress response curve (across time) to estimate repeatability for individuals.
‘profrep’ is a R package for computing profile repeatability on any number of individuals, any number of timepoints, and any number of replicate stress response curves. A full explanation of the math behind Profile Repeatability can be found in Reed et al. (2019).
You can install the development version of profrep from GitHub using devtools with:
You can install the stable version of profrep from CRAN with:
Alternatively, if one is using the use_this package, profrep can be installed with:
The most common use pattern for profrep is to load in your data as a data frame to the active session, and pass it to the main profrep function. Below, we load in an example data set provided with the profrep package:
library(profrep)
my_data <- profrep::synthetic_data_four_point
n_trials <- 4 # or however many trials/rows of data per individual exist
profrep::profrep(df=my_data, n_timepoints=n_trials)
#> individual n_crossings max_variance ave_variance base_score final_score rank
#> 1 E 0 6.67 5.42 12.10 0.9925 1
#> 2 B 0 15.00 12.92 27.95 0.9912 2
#> 3 D 0 26.67 26.40 53.12 0.9887 3
#> 4 F 0 58.92 56.59 115.62 0.9790 4
#> 5 G 0 91.67 87.42 179.27 0.9611 5
#> 6 I 0 106.67 106.67 213.55 0.9461 6
#> 7 J 5 106.67 106.67 277.33 0.9026 7
#> 8 C 0 207.58 86.81 294.81 0.8861 8
#> 9 K 15 106.67 106.67 384.00 0.7613 9
#> 10 H 0 375.00 149.42 525.17 0.4374 10
#> 11 A 0 456.25 181.88 639.04 0.1993 11If you use profrep in your own published work, we ask that you include a reference both to the original paper describing the method (Reed et al., 2019) and the paper introducing this package (Beattie et al., in prep.)