Test the association between two categories

Cross-tabulates two categorical variables and tests whether they are associated.

Usage

Analyze_Contingency(
  dfData,
  strRowCol,
  strColCol,
  strMethod = "chisq",
  chrRowGroups = NULL,
  chrColGroups = NULL,
  nConfLevel = 0.95,
  nMinGroup = nMinGroupDefault
)

Arguments

dfData

data.frame One row per participant.

strRowCol, strColCol

character Names of the two categorical columns: the rows and the columns of the table.

strMethod

character The test: "chisq" or "fisher". Default: "chisq".

chrRowGroups, chrColGroups

character The categories to keep, in order, for the rows and for the columns. Participants in any other category are dropped and counted. On a two-by-two table the order sets which way round the odds ratio is. Default: NULL, every category present, in sorted order.

nConfLevel

numeric Confidence level of the odds ratio's interval. Default: 0.95.

nMinGroup

numeric The smallest category the test is computed for. If any row or column of the table totals fewer participants, the result has status "too_small" and no numbers. Default: nMinGroupDefault, which is 5. See StatisticsResult.

Value

The fixed result described in StatisticsResult. Here counts is the number of participants in the table; estimates has the odds ratio for "fisher" on a two-by-two table and no rows otherwise; and rows is the table in long form, one row per cell, with the columns row, col, n, expected and small_expected (the last two NA for "fisher").

Details

Each method is the base R function, called with R's defaults:

strMethod	R function
"chisq"	stats::chisq.test(), with the continuity correction on a two-by-two table.
"fisher"	stats::fisher.test(), on a table of any size R will accept. On a two-by-two table it also estimates the odds ratio, with its interval.

For "chisq", each cell of rows carries the expected count that chisq.test() returned and is flagged where it is below 5, the count at which chisq.test() itself warns. When any cell is flagged, notes says how many, and R's warning is in warnings.

See also

Other statistics: Analyze_Correlation(), Analyze_CorrelationMatrix(), Analyze_Fit(), Analyze_GroupDifference(), Analyze_Screen(), Analyze_Survival()

Examples

lResult <- Analyze_Contingency(Synthetic_Participants, "ARM", "RESPONSE")
lResult$method
#> [1] "Pearson's Chi-squared test with Yates' continuity correction"
lResult$p_value
#> [1] 0.4639908
lResult$rows
#>         row           col  n expected small_expected
#> 1   Placebo Non-responder 66       63          FALSE
#> 2 Treatment Non-responder 60       63          FALSE
#> 3   Placebo     Responder 34       37          FALSE
#> 4 Treatment     Responder 40       37          FALSE

Analyze_Contingency(Synthetic_Participants, "ARM", "RESPONSE", strMethod = "fisher")$estimates
#>         name group estimate     lower   upper level
#> 1 odds ratio  <NA> 1.292434 0.6994128 2.39815  0.95