Start here

statsguider is built around one simple idea:

1. describe the data properties
2. let the package choose the test
3. run the recommended method

Main arguments in select_test()

There are two kinds of arguments.

Columns in your table:

• data: the full data.frame
• outcome: the column you want to analyze
• group: the column that defines groups or conditions
• id: the subject ID column for paired or repeated data

Choice-style arguments:

• goal
  • "difference": compare groups
  • "association": measure association
  • "adjusted_effect": estimate an adjusted effect
  • "time_to_event": analyze time-to-event data
  • "agreement": measure agreement or reproducibility
  • "equivalence": test equivalence or non-inferiority
• outcome_type
  • "continuous": numeric values
  • "binary": yes/no outcomes
  • "nominal": unordered categories
  • "ordinal": ordered categories
  • "count": event counts
• paired
  • "yes": the same subject is measured twice
  • "no": different subjects are compared
• repeated
  • "yes": the same subject is measured three or more times
  • "no": not repeated-measures data
• adjust
  • "yes": covariate adjustment is needed
  • "no": a simple unadjusted comparison is enough
• normality
  • "auto": let statsguider check automatically
  • "yes": use the normal-data branch
  • "no": use the non-normal branch
  • "unknown": you are not sure
• run
  • "recommend": choose a method only
  • "run": choose and run the method
• language
  • "en": English output
  • "ja": Japanese output

How to set them in practice

• Use goal = "difference" for ordinary group comparison.
• Use outcome_type = "continuous" for numeric lab values.
• Use paired = "yes" when the same subject appears twice.
• Use repeated = "yes" when the same subject appears at three or more visits or conditions.
• Use adjust = "yes" when your main analysis needs covariates.
• Use run = "recommend" first if you only want the method.

The simplest function

dat <- subset(statsguider::wet_example, visit == "week4")

select_test(
  data = dat,
  outcome = "biomarker",
  group = "group",
  outcome_type = "continuous",
  paired = "no",
  repeated = "no",
  run = "recommend"
)
#> statsguider decision
#> - Action: recommend
#> - Recommended method: Welch t-test
#> - Alternative: Mann-Whitney U test
#> - Reason: The data look like two independent groups with a continuous outcome and acceptable normality.
#> - Next step: Run Welch t-test. Use the rank-based alternative if needed.
#> - Notes:
#>   * Normality was checked automatically and classified as `yes`.

In this example:

• data = dat: use this table
• outcome = "biomarker": analyze the biomarker column
• group = "group": compare the groups in group
• outcome_type = "continuous": the outcome is numeric
• paired = "no": the groups are independent
• repeated = "no": this is not repeated-measures data
• run = "recommend": only show the recommended method

Run the test immediately

select_test(
  data = dat,
  outcome = "biomarker",
  group = "group",
  outcome_type = "continuous",
  paired = "no",
  repeated = "no",
  run = "run"
)
#> statsguider result
#> - Method: Welch t-test
#> - Reason: The data look like two independent groups with a continuous outcome and acceptable normality.
#> - Summary: Welch t-test was selected because the data looked like continuous outcome, 2 groups, paired = "no", repeated = "no".

The only important change here is run = "run".

guided_test() uses the same choices

guided_test(
  dat,
  answers = list(
    goal = "difference",
    outcome = "biomarker",
    group = "group",
    paired = "no",
    repeated = "no",
    adjust = "no",
    outcome_type = "continuous",
    normality = "auto"
  ),
  language = "en"
)
#> statsguider decision
#> - Action: recommend
#> - Recommended method: Welch t-test
#> - Alternative: Mann-Whitney U test
#> - Reason: The data look like two independent groups with a continuous outcome and acceptable normality.
#> - Next step: Run Welch t-test. Use the rank-based alternative if needed.
#> - Notes:
#>   * Normality was checked automatically and classified as `yes`.

guided_test() uses the same logic as select_test(). The difference is only the interface:

• select_test(): set the arguments directly
• guided_test(): answer the questions one by one