Tidy a 'meta' object into a plottable data frame

Converts objects of class meta (from the meta package) into a tidy data frame suitable for use with ggmeta geometries and ggforest(). The returned data frame has one row per study or summary.

Usage

tidy_meta(x, ...)

# Default S3 method
tidy_meta(x, ...)

# S3 method for class 'data.frame'
tidy_meta(
  x,
  add_summary = FALSE,
  summary_method = c("common", "random"),
  level = 0.95,
  ...
)

# S3 method for class 'meta'
tidy_meta(
  x,
  back_trans = c("auto", "exp", "none"),
  sort_studies = TRUE,
  add_summary = TRUE,
  add_predict = TRUE,
  add_subgroups = TRUE,
  ...
)

Arguments

x: An object of class meta, e.g. created by meta::metabin(), meta::metacont(), or meta::metagen().
...: Additional arguments passed to methods.
add_summary: For the data-frame method, if TRUE compute an inverse-variance / DerSimonian-Laird pooled summary from the study rows and append it (on-the-fly meta-analysis, no meta package required). Needs a se column, or ci_lower/ci_upper to recover it. Default FALSE.
summary_method: Which pooled summaries to add when add_summary = TRUE: "common", "random", or both (default).
level: Confidence level for the pooled summary interval. Default 0.95.
back_trans: Should the estimate and confidence limits be back-transformed to the natural scale? If "auto" (default), each summary measure is back-transformed with its correct inverse via meta::backtransf() (exponentiation for ratios, inverse-logit for PLOGIT, Fisher's z to correlation for ZCOR, etc.); linear measures are left unchanged. Use "exp" to force exponentiation or "none" to keep the analysis scale.
sort_studies: If TRUE (default), sort studies by effect estimate (most favorable at top).
add_predict: If TRUE (default), include prediction interval row when available.
add_subgroups: If TRUE (default), include subgroup headers and within-group summary rows when subgroups are present.

Value

A data.frame with one row per study or summary, with columns:

studlab: Study label (character)
estimate: Point estimate (numeric)
ci_lower: Lower confidence limit (numeric)
ci_upper: Upper confidence limit (numeric)
se: Standard error (numeric)
weight: Study weight (numeric, NA for summaries)
p_value: P-value (numeric)
n: Sample size or person-time (numeric, optional)
event: Number of events (numeric, optional)
is_summary: Logical, TRUE for summary rows
summary_type: Character: "none", "common", "random", "subgroup", or "predict"
subgroup: Subgroup label (character or NA)

The returned data frame has the following attributes:

sm: summary measure type (e.g. "RR", "OR", "MD")
null_effect: null effect value for reference line
method: meta-analysis method
common: logical, TRUE if common effect model was used
random: logical, TRUE if random effects model was used
tau: heterogeneity estimate tau
k: number of studies

Examples

# \donttest{
library(meta)
m <- metabin(event.e, n.e, event.c, n.c,
  data = data.frame(
    event.e = c(14, 30), n.e = c(100, 150),
    event.c = c(10, 25), n.c = c(100, 150)
  ),
  studlab = c("Study A", "Study B"),
  sm = "RR"
)
tidy_meta(m)
#>                    studlab estimate   ci_lower  ci_upper        se    weight
#> 1                  Study A 1.400000 0.65296960  3.001671 0.3891382  6.603774
#> 2                  Study B 1.200000 0.74247238  1.939466 0.2449490 16.666667
#> common       Common effect 1.257143 0.83734105  1.887413 0.2073331        NA
#> random      Random effects 1.253660 0.83507644  1.882058 0.2072992        NA
#> 11     Prediction interval 1.253660 0.09000235 17.462462 0.2072992        NA
#>          p_value is_summary summary_type subgroup
#> 1      0.3872255      FALSE         none     <NA>
#> 2      0.4566801      FALSE         none     <NA>
#> common 0.2697065       TRUE       common     <NA>
#> random 0.2754777       TRUE       random     <NA>
#> 11            NA       TRUE      predict     <NA>
# }