ratings_vs_bench: Compare ratings to a benchmark
In darrellpenta/uxstats: Common data analysis tools for UX researchers

Description Usage Arguments Details Value See Also Examples

View source: R/ratings_vs_bench-function.R

ratings_vs_bench() tests a sample mean of continuous data (e.g., ratings) against a given benchmark using a one-sample t-test.

ratings_vs_bench(.x, ...)

## S3 method for class 'numeric'
ratings_vs_bench(
  .x,
  .sd,
  .n,
  .m,
  ...,
  .alt = c("greater", "less", "twotailed"),
  .alpha = 0.05
)

## S3 method for class 'data.frame'
ratings_vs_bench(.x, .var, .m, ..., .alt = "greater", .alpha = 0.05)

`.x`	A single numeric value, a vector of values, or a long-format data frame with a named column of numeric data corresponding to ratings or another continuous metric. See Details.
`...`	(Optional) If `.x` is a long-format data frame, you can pass the name of one or more grouping variables as unquoted, comma-separated column names (without naming the `...` argument) to compute stats by groups.
`.sd`	If `.x` is a single numeric value, `.sd` should be a single numeric value indicating the sample standard deviation. See Details
`.n`	If `.x` is a single numeric value, `.n` should be a single numeric value indicating the sample size. See Details.
`.m`	The test (benchmark) mean.
`.alt`	(Optional) For test alternatives, one of `c("greater","less","twotailed")`. Defaults to "greater" for a one-sided test.
`.alpha`	(Optional) A positive number (where 0 < `.alpha` < 1) specifying the significance level to be used. Defaults to `.alpha = 0.05`. To set a different significance level, the argument must be named (i.e., `.alpha=0.001`) or else the function may yield unexpected results.
`.var`	If `.x` is a long-format data frame, the (unquoted) name of a data frame column containing task ratings (or values from another continuous metric).

ratings_vs_bench assumes that you want to test the hypothesis that the observed outcome exceeds the benchmark, and therefore, defaults to a one-tailed test. This means that setting .alpha = 0.05 (the default) produces a 90% confidence interval.

If .x is a single numeric value representing the sample mean, provide the sample standard deviation and sample size to .sd and .n, respectively.
If .x is a numeric vector of ratings values, you should only specify the test (benchmark) mean.
If .x is a data frame, .var should be the unquoted name of the column containing the ratings values.
If you're passing a data frame to .x, you can optionally pass one or more grouping variables as unquoted, comma-separated column names (without naming the ... argument) to compute stats by groups.
You can choose from among the test alternatives c("greater","less","twotailed") by providing one of the options to the .alt argument: e.g., .alt = "twotailed". Defaults to "greater" for a one-sided test.
You can modify the alpha level to adjust confidence intervals by including .alpha as a named argument and providing a numeric value: e.g., .aplha = 0.001.

Note that NAs are automatically dropped in all calculations.

A tibble with data summaries and test results

Other benchmark comparison stats: success_vs_bench(), time_vs_bench()

# Comparing a sample ratings mean of 79 (sd = 20) from 160 users
# against a benchmark mean of 75
ratings_vs_bench(.x=79,.sd=20,.n=160,.m=75)

# Comparing values from a data frame against a benchmark of 70.
.ux_data <-
 data.frame(
  "id" = rep(seq(1,10,1),2),
  "task" = c(rep(1,10),rep(2,10)),
  "ratings"  = runif(20,0,100))

ratings_vs_bench(.ux_data, ratings, .m=70, task)