Coded relational values and disvalues of landowners adjacent to Cerro Castillo National Park, Chilean Patagonia: analysis dataset and reproducible R code (2024)

Published: 6 September 2026| Version 2 | DOI: 10.17632/shnyvb77kg.2
Contributors:
,
,

Description

This dataset supports an exploratory study of how relational valuation contributes to understanding people–nature relationships among landowners adjacent to Cerro Castillo National Park, Chilean Patagonia, in a context of intensifying rural land subdivision. It contains: (i) a de-identified analysis dataset of 129 landowners who completed a relational-values module, comprising 21 categorical variables covering property patterns, sociodemographic characteristics, and coded relational values and disvalues; (ii) an aggregated sample-comparison dataset (n = 641) used to compare module respondents with the full survey sample; (iii) a redacted coding audit trail documenting the classification of each open-ended response; and (iv) a single R script reproducing all reported analyses, including Multiple Correspondence Analysis, bootstrap stability assessment, Hierarchical Clustering on Principal Components, and the comparison of the two profiling models. Data were derived from the Phase 2 Survey of ANID Fondecyt 1230020 (2023–2026). An open-access, non-identifiable version of that survey dataset is available at https://doi.org/10.17632/xz3fpvtb7n.1. Verbatim open-ended responses are not included, in order to protect participant confidentiality. Personal names, locality names and identifiers have been masked in the quotations reported in the coding audit trail. Users must not attempt to re-identify individual respondents.

Files

Steps to reproduce

All results reported in the associated article are reproduced by a single R script from the three CSV files in this deposit. No other data are required. 1. SOFTWARE Install R 4.3 or later (the analysis was originally run under R 4.5.0), then install the required packages: install.packages(c("tidyverse", "FactoMineR", "factoextra", "missMDA", "readxl", "writexl")) 2. SETUP Place these four files in the same folder and set it as the R working directory: Supplementary_Code_S5_Reproducible_Analysis.R Supplementary_Data_S6_Analysis_Dataset.csv Supplementary_Data_S7_Sample_Comparison.csv Supplementary_Data_S8_Coding_Audit_Trail.csv The CSVs are UTF-8 with a byte-order mark. Do not re-save them from Excel: in some locales Excel writes CSVs with a semicolon separator, which corrupts the file. Open read-only or work from a copy. 3. RUN source("Supplementary_Code_S5_Reproducible_Analysis.R", encoding = "UTF-8") Open and save the script as UTF-8. The run switches at the top are preset to the published values and need no changing. BUILD_PUBLIC_DATASET must remain FALSE; it rebuilds the analysis dataset from internal project files that are not public and is included only to document how this deposit was derived. The full run takes roughly ten minutes, almost all of it in the two bootstrap loops (1,000 replications each). All outputs are written to an Output/ subfolder. 4. WHAT THE SCRIPT DOES Loads the analysis dataset (n = 129); reproduces Table 4 from the sample-comparison file (n = 641); imputes two missing values by MCA-based single imputation (missMDA, 2 dimensions, seed 123); reproduces Tables 5 and A3; runs the preliminary MCA for variable screening (Supplementary Table S2); then estimates both models, each with Benzecri-corrected inertia, bootstrap stability of category coordinates, and HCPC. Model A retains 2 dimensions, Model B retains 4; both are cut at 3 clusters. Finally it compares the two partitions by contingency table and chi-square. 5. EXPECTED OUTPUT 23 files in Output/, all prefixed by model label. Key values to verify: Model A clusters: 53 / 43 / 33; first two dimensions 33.63% of raw inertia; Benzecri Dim1 87.49%, Dim2 9.78% Model B clusters: 28 / 43 / 58; first two dimensions 27.95% of raw inertia; Benzecri Dim1 66.72%, Dim2 22.22%; cumulative raw inertia over four dimensions 47.74% Model A vs Model B: chi-square = 19.886, df = 4, p < 0.001 Cluster solutions, eigenvalues, corrected inertia and the chi-square are deterministic and reproduce exactly. Bootstrap standard deviations for the least stable categories may differ in the second decimal place across R and FactoMineR versions; the classification of each category as stable (SD <= 0.8) or requiring cautious interpretation (SD > 0.8) is unaffected. A full description of the files, variables and confidentiality measures is given in README.docx.

Categories

Social Behavior, Values, Conservation Behavior

Funders

Licence