Dataset DOI: 10.5061/dryad.gf1vhhn54

Description of the data and file structure

This dataset supports a meta-analysis of how silicon (Si) supplementation affects plant stoichiometry. We searched Web of Science Core Collection on 19 October 2024, screened records in Covidence, and extracted data from 168 primary studies that compared Si-supplemented plants (treatment) with plants without added Si (control). Each row of the dataset is one treatment–control comparison for one response variable: plant Si, C, N or P concentration, C:N or C:P ratio, or N or P use efficiency (NUE, PUE). The effect size is the log response ratio (lnRR). A positive lnRR means the value was higher in Si-supplemented plants

Files and variables

File: dataset.csv

Description: Final dataset used in all analyses (1,532 rows, 30 columns)

Variables

• StudyID: Study identifier: Covidence number, first author and year (e.g., "#167Akca 2023") | text

• ObsID: Unique identifier of each effect size (not consecutive) | integer

• Observation_within_study: Number of the treatment–control comparison within a study | integer

• Published_Year: Year of publication | year

• Species: Plant species (currently accepted name) | text

• Family: Plant family | text

• Shoot_Root: Tissue measured | "Shoot" (aboveground tissue), "Root"

• Poaceae_or_Non: Grass family or not | "Poaceae", "Non-Poaceae"

• Si_type: Form of Si applied | "Silicate", "Silicic acid", "Nano-Si" (nanoparticles), "Silica" (SiO2), "Diatomite" (diatomaceous earth), "Bio-Si" (biogenic silica)

• Control_samplesize:  Number of replicates, control | count

• Treatment_samplesize:  Number of replicates, Si treatment | count

• Growth_condi: Growth condition | "Hydroponic", "Pot", "Field"

• Response: Response variable | "Si", "C", "N", "P" (concentration); "CN" (C:N ratio); "CP" (C:P ratio); "NUE" (N use efficiency); "PUE" (P use efficiency)

• Control_mean:  Mean, control | see units below

• Control_se: Standard error, control | as mean 

• Treatment_mean: Mean, Si treatment | see units below

• Treatment_se:  Standard error, Si treatment | as mean

• Control_sd: Standard deviation, control (SE × √n) | as mean

• Treatment_sd:  Standard deviation, Si treatment (SE × √n) | as mean

• yi: Log response ratio ln(Treatment_mean / Control_mean), from metafor::escalc; for reference only | unitless

• vi: Sampling variance of yi; for reference only | unitless

• Unique_ID: StudyID and Observation_within_study joined with "_" | text

• cv_control:  Coefficient of variation, control (SD / mean) | unitless

• cv_treatment: Coefficient of variation, Si treatment (SD / mean) | unitless

• b_CV2_2:  Weighted average squared CV across studies, control; calculated per response (NUE and PUE pooled) and used to fill missing CVs | unitless

• b_CV2_1:  As b_CV2_2, Si treatment | unitless

• cv2_control_new:  Squared CV used for the control: cv_control² if available, otherwise b_CV2_2 | unitless

• cv2_treatment_new: Squared CV used for the Si treatment: cv_treatment² if available, otherwise b_CV2_1 | unitless

• lnRR_new:  Effect size used in all models: bias-corrected lnRR (Lajeunesse 2015) | unitless 

• lnRRV_new: Sampling variance of lnRR_new, used in all models | unitless

  NOTE

• Units of the response variables:

  - Si, C, N, P: % of dry mass. Values reported in other units were converted to %.

  - CN, CP: ratios. Where a study did not report them, they were calculated as %C/%N and %C/%P.

  - NUE, PUE: units as reported in each study. The effect sizes (lnRR) are unitless and comparable across studies.

• `NA` in the SE, SD, CV and vi columns (558 control rows, 559 treatment rows) means the primary study did not report variation. For these rows, the missing CV was replaced by the weighted average CV (b_CV2_1, b_CV2_2), following Nakagawa et al. (2023), so lnRR_new and lnRRV_new have no missing values.

File: Script_outputPDF.Rmd

Description: R Markdown script that reproduces all analyses, figures and supplementary tables

File: Excluded_records.csv

Description: Records excluded at the full-text screening stage, with reasons (284 rows, 10 columns)

Variables

• Covidence ID: Covidence number, first author and year

• Title:  Article title

• Authors:  Authors, separated by ";"

• Published Year: Year | Year of publication

• Journal: Journal name

• Volume: Journal volume

• Issue: Journal issue

• Pages: Page range or article number

• DOI: Digital Object Identifier

• Exclusion reason:  "No wanted response" (none of the target variables reported), "No control" (no treatment without Si), "Not Si manipulation" (Si not experimentally manipulated), "Foliage Si application" (Si applied to leaves), "Review paper", "Abstract only" (full text not available), "Soil study" (only soil measured), "No sample size" (sample size not reported)

  NOTE

  `NA` in Volume, Issue, Pages or DOI means the record has no such information

Code/software

Script_outputPDF.Rmd runs in R using RStudio platform (R 4.5.1).

Main R packages: metafor, orchaRd (install from GitHub: daniel1noble/orchaRd), clubSandwich, MuMIn, tidyverse, knitr, kableExtra, patchwork, ggpubr. The script loads all packages with pacman::p_load.

Access information

Data was derived from the following sources:

• Data were extracted from the published studies listed by StudyID (see the reference list in the article's Supporting Information).