Skip to contents

Correlate two numeric variables

Source: R/statistics.R
Analyze_Correlation.Rd

Estimates the correlation between two numeric variables on the participants who have both, overall and, when a group column is named, within each group.

Usage

Analyze_Correlation(
  dfData,
  strXCol,
  strYCol,
  strMethod = "pearson",
  strGroupCol = NULL,
  chrGroups = NULL,
  nConfLevel = 0.95,
  nMinGroup = nMinGroupDefault
)

Arguments

dfData

data.frame One row per participant.

strXCol, strYCol

character Names of the two numeric columns.

strMethod

character The coefficient: "pearson" or "spearman". Default: "pearson".

strGroupCol

character Name of a column holding each participant's group, to add a row per group. Default: NULL, no per-group rows.

chrGroups

character The groups to give a row to, in order. Default: NULL, every group present, in sorted order.

nConfLevel

numeric Confidence level of the intervals. Default: 0.95.

nMinGroup

numeric The smallest number of complete pairs the correlation is computed for, overall and in each group. Default: nMinGroupDefault, which is 5. See StatisticsResult.

Value

The fixed result described in StatisticsResult. Here counts is the number of complete pairs; estimates has one row, named as R names the coefficient ("cor" or "rho"); and rows has one row per group when strGroupCol is given, with the columns group, counts, estimate, lower, upper, level, method, statistic, p_value, adjustment, status, reason and warning. The overall answer uses every complete pair, whether or not the participant has a group.

Details

Both methods are stats::cor.test() with R's defaults. For "pearson" the interval is the one cor.test() gives, by Fisher's z. For "spearman" cor.test() gives no interval, so none is reported and notes says so.

See also

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

Examples

# TNF-alpha against IL-10 at Baseline
dfBaseline <- Synthetic_Results[Synthetic_Results$VISIT == "Baseline", ]
dfTNF <- dfBaseline[dfBaseline$TEST == "TNF-alpha", ]
dfIL10 <- dfBaseline[dfBaseline$TEST == "IL-10", ]
dfFrame <- Synthetic_Participants
dfFrame$TNF <- dfTNF$STRESN[match(dfFrame$USUBJID, dfTNF$USUBJID)]
dfFrame$IL10 <- dfIL10$STRESN[match(dfFrame$USUBJID, dfIL10$USUBJID)]

lResult <- Analyze_Correlation(dfFrame, "TNF", "IL10", strGroupCol = "ARM")
lResult$estimates
#>   name group  estimate     lower     upper level
#> 1  cor  <NA> 0.6383823 0.5481947 0.7138938  0.95
lResult$rows[c("group", "counts", "estimate", "lower", "upper")]
#>       group counts  estimate     lower     upper
#> 1   Placebo    100 0.5918125 0.4474016 0.7061463
#> 2 Treatment    100 0.6736870 0.5500544 0.7684239