get.stats: get.stats

Description

Extracts statistical results from text string, XML, CERMXML, HTML or DOCX files. The result is a list with a vector containing all identified sticked results and a matrix containing the reported standard statistics and recalculated p-values if computation is possible.

Usage

get.stats(
  x,
  output = "both",
  stats.mode = "all",
  recalculate.p = TRUE,
  checkP = FALSE,
  alpha = 0.05,
  criticalDif = 0.02,
  alternative = "undirected",
  estimateZ = FALSE,
  T2t = FALSE,
  R2r = FALSE,
  select = NULL,
  rm.na.col = TRUE,
  cermine = FALSE,
  warnings = TRUE
)

Value

If output="all": list with two elements. E1: vector of extracted results by allStats and E2: matrix of standard results by standardStats.
If output="allStats": vector of extracted results by allStats.
If output="standardStats": matrix of standard results by standardStats.

Arguments

x: NISO-JATS coded XML or DOCX file path or plain textual content.
output: Select the desired output. One of c("both", "allStats", "standardStats").
stats.mode: Select a subset of test results by p-value checkability for output. One of: c("all", "checkable", "computable", "uncomputable").
recalculate.p: Logical. If TRUE recalculates p-values of test results if possible.
checkP: Logical. If TRUE observed and recalculated p-values are checked for consistency.
alpha: Numeric. Defines the alpha level to be used for error assignment.
criticalDif: Numeric. Sets the absolute maximum difference in reported and recalculated p-values for error detection.
alternative: Character. Select test sidedness for recomputation of p-values from t-, r- and beta-values. One of c("undirected", "directed"). If "directed" is specified, p-values for directed null-hypothesis are added to the table but still require a manual inspection on consistency of the direction.
estimateZ: Logical. If TRUE detected beta-/d-value is divided by reported standard error "SE" to estimate Z-value ("Zest") for observed beta/d and recompute p-value. Note: This is only valid, if Gauss-Marcov assumptions are met and a sufficiently large sample size is used. If a Z- or t-value is detected in a report of a beta-/d-coefficient with SE, no estimation will be performed, although set to TRUE.
T2t: Logical. If TRUE capital letter T is treated as t-statistic.
R2r: Logical. If TRUE capital letter R is treated as correlation.
select: Select specific standard statistics only (e.g.: c("t", "F", "Chi2")).
rm.na.col: Logical. If TRUE removes all columns with only NA from standardStats.
cermine: Logical. If TRUE CERMINE specific letter conversion will be peformed on allStats results.
warnings: Logical. If FALSE warning messages are omitted.

References

Böschen (2021). "Evaluation of JATSdecoder as an automated text extraction tool for statistical results in scientific reports.” Scientific Reports. doi: 10.1038/s41598-021-98782-3.

Examples

Run this code

## Extract results from plain text input
x<-c("The mean difference of scale A was significant (beta=12.9, t(18)=2.5, p<.05).",
"The ANOVA yielded significant results on 
 faktor A (F(2,18)=6, p<.05, eta(g)2<-.22)",
"the correlation of x and y was r=.37.")
get.stats(x)

## Extract results from native NISO-JATS XML file
# download example XML file via URL if a connection is possible
x<-"https://journals.plos.org/plosone/article/file?id=10.1371/journal.pone.0114876&type=manuscript"
# file name
file<-paste0(tempdir(),"/file.xml")
# download URL as "file.xml" in tempdir() if a connection is possible
tryCatch({
  readLines(x,n=1)
  download.file(x,file)
  },
  warning = function(w) message(
  "Something went wrong. Check your internet connection and the link address."),
  error = function(e) message(
  "Something went wrong. Check your internet connection and the link address.")
)
# apply get.stats() to file
if(file.exists(file)) get.stats(file)