Skip to contents

Import data from a file and return a data frame

Source: R/prepare_data.R
prepare_data.Rd

You may want to use the metafor::escalc() function in the metafor package to calculate effect sizes and their variances in preparation for this function.

Usage

prepare_data(
  data,
  study_label,
  es_field,
  se = NULL,
  pvalue = NULL,
  sample_size = NULL,
  variance = NULL,
  filters = character(),
  url = NA,
  es_type = "SMD",
  article_label = NA,
  es_label = NA,
  na.rm = "es_related",
  arrange_filters = c("given", "alphabetical", "leave"),
  keep_missing_level = FALSE,
  es_id = NULL,
  variance_scale = NULL,
  direction = c("unspecified", "positive", "negative"),
  categorical_filters = character(),
  alternatives = list(),
  primary_label = "As supplied"
)

Arguments

data

path to the .csv file to read OR a data frame

study_label

Character. Name of the field to use as the id/study label

es_field

Character. Name of the field to use as the effect size

se

Character. Name of the field to use as the standard error for the effect size.

pvalue

Character. Name of the field to use as the p-value for the effect size.

sample_size

Character. Name of the field to use as the sample size

variance

Character. Name of the field to use as the sampling variances.

filters

Character. List of fields to use as filters - can be named if different labels should be displayed

url

Character. Field with URLs or DOIs to link to. DOIs can be in the format "10.1234/5678" or full links. Defaults to NA.

es_type

Character. Input scale: SMD, COR (raw r), or ZCOR (Fisher z). COR needs r-scale variance and uses delta-method variance conversion; ZCOR needs z-scale variance.

article_label

Character. Field with article labels. Only used to report number of references in addition to number of independent samples.

es_label

Character. Label for individual effect sizes - only needed when there are multiple effect sizes per study/sample. Defaults to NA. Defaults to NA, in that case, multiple effect sizes are simply numbered.

na.rm

Should rows with any missing values be removed? Can be TRUE, FALSE or "es_related" - the last is the default and drops rows with missing values for any of the variables used in the standard meta-analysis models, namely study_label, es_field, variance, and se. Optional p/N never remove rows under this default. Setting this to TRUE also drops rows with missing values on any of the filters etc, which might often be unnecessary. Conversely, setting this to FALSE might lead to issues in the model - unless you post-process the data or change the models and analyses to be included in the app. Invalid required inputs with FALSE cause an error.

arrange_filters

Character. How should the filters be arranged in the app? Options are "given" by the filters argument, "alphabetical" or "leave" (as they are in the dataset). Defaults to "given".

keep_missing_level

Logical. Should a (Missing)-level be kept even for filters that do not have missing values? Might be advisable when you expect users to upload new data with missing values. Defaults to FALSE.

es_id

Optional column with unique effect IDs within each study. Defaults to input row IDs.

variance_scale

Required for COR (r) and ZCOR (z); optional SMD for SMD.

direction

Explicit direction for one-sided bias models: unspecified, positive, or negative.

categorical_filters

Explicitly convert named filter columns to factors with radix-sorted levels. Useful for CSV/JSON builds.

alternatives

Named list of up to five author-defined effect computations. Each entry supplies es_field, variance or se, es_type, variance_scale when required, and a nonempty justification. Every computation must retain the same effects and converge on the same fitting scale.

primary_label

Label for the primary computation when alternatives are provided.

Value

tibble with the data from the file/input reformatted for metaUI

Examples

raw <- data.frame(study = letters[1:4], d = c(.1, .3, -.1, .4), vi = rep(.02, 4))
prepare_data(raw, "study", "d", variance = "vi")
#> # A tibble: 4 × 14
#>   metaUI__study_id metaUI__effect_size metaUI__variance metaUI__se
#>   <chr>                          <dbl>            <dbl>      <dbl>
#> 1 a                                0.1             0.02      0.141
#> 2 b                                0.3             0.02      0.141
#> 3 c                               -0.1             0.02      0.141
#> 4 d                                0.4             0.02      0.141
#> # ℹ 10 more variables: metaUI__pvalue <dbl>, metaUI__N <dbl>,
#> #   metaUI__es_type <chr>, metaUI__es_label <int>, metaUI__effect_id <int>,
#> #   metaUI__input_effect <dbl>, metaUI__input_variance <dbl>,
#> #   metaUI__display_scale <chr>, metaUI__direction <chr>, metaUI__es_z <dbl>