This README file regards the Dryad distribution for:
Haller, B.C., and Hendry, A.P. (2013). Solving the paradox of stasis: Squashed stabilizing selection and the limits of detection. Evolution.
Corresponding author:
Ben Haller
benjamin.haller@mail.mcgill.ca
Department of Biology and Redpath Museum
McGill University
859 Sherbrooke St. West
Montreal, Quebec H3A 0C4
Canada
------------------------------------------------------------------------------------------------------------------
Seven files are included in this distribution:
Filename Description
------------------ --------------------
realizations.csv a comma-separated values (CSV) file with summary statistics for each realization of the model
gradients.csv a CSV file with histograms of the selection gradients for each realization (see notes below)
generation_data_4595.csv per-generation values for realization 4595 (see notes below)
selective_deaths_4595.csv per-generation selective deaths for realization 4595 (see notes below)
generation_data_4631.csv per-generation values for realization 4631 (see notes below)
selective_deaths_4631.csv per-generation selective deaths for realization 4631 (see notes below)
README.txt this file
------------------------------------------------------------------------------------------------------------------
The realizations.csv file contains a bunch of rows plus an initial header row. Each row contains values for the following 48 columns. The columns are reordered and divided into sections here, to make it more clear where they are used in the published paper.
CORE COLUMNS (used in main paper analyses):
Name Description
------------------ --------------------
realization 1...12096, an index for the realization
Nj 1000 (main analyses) or 500/2500 (supplemental analyses for small/large pop. size)
XCp NO or YES: competition (i.e. negative freq.-dep. selection) on?
geneticsVariant "Q", "T", or "C": the genetic architecture used (quantitative, triallelic, continuum)
VE 0.001, 0.01, 0.1: the level of environmental variance used
mu 1e-05 or 0.001: the probability of mutation, mu
omega 1 or 10: the width of the stabilizing selection function (1==strong selection, 10==weak)
sigma_c 0.5 or 2.0: the width of the competition function, sigma_c
m 0.0, 0.1, or 0.5: the probability of random mortality
alpha 0.05, 0.5, or 1.0: the mutational effect size (1.0 for triallelic architecture realizations)
subset_size 100, 500, 1000, or 2500: the mark-recapture subset size taken for this sampling of the realization
analysis_var "a1", "an", "z1", "zn": "a" means genotype, "z" means phenotype, "1" is selected ("s" in the paper), "n" is neutral
linreg1_l_mean_sig P(β*), the rate of detection of linear selection; the dependent variable in Figure 3
linreg2_q_mean_sig P(γ*), the rate of detection of quadratic selection; the dependent variable in Figure 4
linreg2_mean_sig_gamma_sign the median proportion of detected quadratic selection that was stabilizing (see Figure 4 caption)
MUTATIONAL VARIANCE AND HERITABILITY, SUPPLEMENTAL FIGURE S2.1 and S2.2:
Name Description
------------------ --------------------
VM mutational variance, calculated as given in the paper; independent variable for Figures S2.1 and S2.2ab
selected_trait_mean_h2 mean heritability of the selected trait (calculated from the full population, not the mark-recapture subsample)
neutral_trait_mean_h2 mean heritability of the neutral trait (calculated from the full population, not the mark-recapture subsample)
SELECTIVE MORTALITY, SUPPLEMENTAL FIGURE S2.3:
Name Description
------------------ --------------------
n_died_s selective mortality rate (the mean proportion of individuals that died due to selection per generation)
LOGISTIC VERSUS LINEAR REGRESSION, SUPPLEMENTAL FIGURE S2.17:
Name Description
------------------ --------------------
logreg1_l_mean_sig P(β*)[logistic], the y axis in Figure S2.17a
logreg2_q_mean_sig P(γ*)[logistic], the y axis in Figure S2.17b
linreg1_median_sig_abs_beta the median significant absolute β, the x axis in Figure S2.17c
logreg1_median_sig_abs_beta_avggrad the median significant absolute β_avggrad, the y axis in Figure S2.17c
linreg2_median_sig_abs_gamma the median significant absolute γ, the x axis in Figure S2.17d
logreg2_median_sig_abs_gamma_avggrad the median significant absolute γ_avggrad, the y axis in Figure S2.17d
INTRINSIC RATE OF EVOLUTION, SUPPLEMENTAL FIGURE S2.18:
Name Description
------------------ --------------------
ir_abs_median the median intrinsic rate of evolution h0; the y axis in Figure S2.18 is the mean of these values
TEMPORAL VARIATION, SUPPLEMENTAL FIGURES S2.19 AND S2.20:
Name Description
------------------ --------------------
linreg1_median_abs_beta the median absolute β, the x axis for Figure S2.19a
linreg2_median_abs_gamma the median absolute γ, the x axis for Figure S2.19b
linreg1_mad_beta the MAD of β, the x axis for Figure S2.19c
linreg1_mad_abs_beta the MAD of absolute β, the x axis for Figure S2.19e
linreg2_mad_gamma the MAD of γ, the x axis for Figure S2.19d
linreg2_mad_abs_gamma the MAD of absolute γ, the x axis for Figure S2.19f
linreg1_median_se_beta the median SE of β, the x axis for Figure S2.20c
linreg2_median_se_gamma the median SE of γ, the x axis for Figure S2.20d
ESTIMATION OF FITNESS LANDSCAPE PARAMETERS, SUPPLEMENTAL FIGURE S2.23:
Name Description
------------------ --------------------
linreg2_median_estimated_omega the median estimated omega, used to calculate the y axis for Figures S2.23ab
linreg2_median_estimated_theta the median estimated theta, used to calculate the y axis for Figures S2.23cd
AUTOCORRELATION AND REPRODUCIBILITY:
Name Description
------------------ --------------------
beta_LB_p probability of significant autocorrelation (Ljung-Box test) for β values
beta_sig_LB_p probability of significant autocorrelation (Ljung-Box test) for significant β values
gamma_LB_p probability of significant autocorrelation (Ljung-Box test) for γ values
gamma_sig_LB_p probability of significant autocorrelation (Ljung-Box test) for significant γ values
beta_first_ac the strength of short-term autocorrelation for β values
beta_sig_first_ac the strength of short-term autocorrelation for significant β values
gamma_first_ac the strength of short-term autocorrelation for γ values
gamma_sig_first_ac the strength of short-term autocorrelation for significant γ values
beta_scl_95 the decay time for autocorrelation for β values
beta_sig_scl_95 the decay time for autocorrelation for significant β values
gamma_scl_95 the decay time for autocorrelation for γ values
gamma_sig_scl_95 the decay time for autocorrelation for significant γ values
------------------------------------------------------------------------------------------------------------------
The gradients.csv file provides histogram (binned) data for the selection gradients obtained in each realization.
There are a bunch of rows, each corresponding to a row in the realizations.csv file; the per-realization values in
realizations.csv can be used to select rows in gradients.csv, since the row numbers correspond. Each row consists of
4001 columns. The first column is the realization number, matching the realization number in realizations.csv.
The rest of the rows are:
2:1001 : a histogram (absolute frequency values) of non-significant beta estimates obtained across all generations
1002:2001 : a histogram (absolute frequency values) of significant beta estimates obtained across all generations
2002:3001 : a histogram (absolute frequency values) of non-significant gamma estimates obtained across all generations
3002:4001 : a histogram (absolute frequency values) of significant gamma estimates obtained across all generations
Each set of 1000 columns spans selection gradient estimates from -1.0 to 1.0, with a step of 0.002; bin midpoints are
thus from -0.999 to 0.999 with a step of 0.002, and there are 1000 bins per histogram. Selection gradient estimates
outside of [-1.0, 1.0] are placed in the first or last bin as appropriate.
These histograms may be used to generate figures essentially equivalent to Figures 5, 6, S2.4-S2.12, and S2.22.
Those figures were generated using the original, unbinned beta and gamma values, so small differences may exist due
to the quantization introduced by binning, but these effects should be minor.
------------------------------------------------------------------------------------------------------------------
Per-generation files. These files give more complete data for two of the realizations, realization #4595, which is
the non-competition "case study" shown in Figure S2.13, and realization #4631, which is the competition "case study"
shown in Figure S2.14. These realizations are also the basis for Figures S2.15 and S2.16.
The "generation_data" CSV files have five columns, with values for each of the 50,000 post-burn-in generations:
Name Description
------------------ --------------------
beta the beta estimate from the non-quadratic linear regression
beta_sig the p-value for that beta estimate
gamma the gamma estimate from the quadratic linear regression
gamma_sig the p-value for that gamma estimate
mean_z the mean phenotypic (z) value for the selected trait
The "selective_deaths" CSV files have a variable number of columns. The first column is always present, and is the
number of the generation, from 1 to 50,000. Remaining columns, if present, are the phenotypic values (z) for the
selected trait for each individual that died in that generation due to selection. (Note there is no random mortality
in these realizations.) Generations in which no selective deaths occurred have no values beyond the generation number.
------------------------------------------------------------------------------------------------------------------
See the paper for more information about these parameters and metrics. Enjoy!