Skip to contents

The R code behind an analysis result

Source: R/analysis_syntax.R
analysis_syntax.Rd

Returns the statistical call that produced a result, as R code a reader can copy and run. This is what lets a report show t.test() or cor.test() with the variables and options resolved, rather than only the run_analysis_plan() call that dispatched it.

Usage

analysis_syntax(x, which = NULL, data_expr = "scored", header = FALSE)

Arguments

x

An sframe_analysis_results object from run_analysis_plan(), or one block's result from it.

which

Character or NULL. One block id, when x holds several.

data_expr

Character. The expression the code should read its data from. Defaults to "scored", the scored frame the header sets up.

header

Logical. Whether to include the lines that load the instrument, read the responses and score the scales. TRUE gives a script that runs on its own.

Value

A character vector of R code lines, or NULL for a method this does not cover. For several blocks, a named list of such vectors.

Details

The code is built from the same resolved specification the runner executed: the variables in the order it resolved them, and the options after defaults were applied. Running it reproduces the statistic, degrees of freedom and p value the package reports, which the package's own tests check by running the generated code and comparing.

What it covers

The 2-group, paired, k-group, correlation, regression and categorical families, and descriptives. sframe_syntax_methods lists them. A method outside that list returns NULL: the model families carry their own syntax already, through cfa_syntax() and its neighbours, and for the rest the computation has no single base-R equivalent to show honestly.

See also

run_analysis_plan(), cfa_syntax(), render_report()

Examples

instr <- sf_instrument("Syntax demo", components = list(
  sf_item("score", "Score", type = "numeric"),
  sf_item("arm", "Arm", type = "text")
))
sf_plan(instr) <- list(list(
  id = "RQ1", research_question = "Do the arms differ?",
  family = "inferential", method = "t_test_ind",
  roles = list(group = "arm", outcome = "score")
))

set.seed(1)
responses <- data.frame(
  arm   = rep(c("control", "treatment"), each = 15),
  score = c(rnorm(15, 10), rnorm(15, 12))
)
results <- run_analysis_plan(responses, instr)

# the call behind the number, with the variables and options resolved
cat(analysis_syntax(results, which = "RQ1"), sep = "\n")
#> # Do the arms differ?
#> # method: t_test_ind | n = 15 + 15
#> group   <- as.character(scored[["arm"]])
#> outcome <- as.numeric(as.character(scored[["score"]]))
#> levels_seen <- unique(group[!is.na(group)])
#> g1 <- outcome[group == levels_seen[1]]; g1 <- g1[!is.na(g1)]
#> g2 <- outcome[group == levels_seen[2]]; g2 <- g2[!is.na(g2)]
#> t.test(g1, g2, var.equal = FALSE)