Calculate complete-case Pearson correlations

Calculate correlations using one common sample, like Stata's corr (correlate). Rows missing any selected variable are excluded from every correlation. Use pwcorr() to use a separate sample for each pair instead.

Usage

corr(
  .data,
  ...,
  .by = NULL,
  .wide = FALSE,
  .wt = NULL,
  .wt_type = c("frequency", "analytic")
)

Arguments

.data: A data frame, tibble, or data.table.
...: <tidy-select> or <data-masking> Columns to select or compute with select_or_transmute(). Existing columns, selection helpers, and named expressions can be mixed. Results must be numeric or logical vector columns. Matrix columns are unsupported. Omit to use every numeric or logical vector column; other columns are then skipped. Selecting another type is an error.
.by: <tidy-select> Optional grouping variables. Grouped data use their groups unless .by is supplied. Select existing columns only; create new grouping columns with dplyr::mutate() first.
.wide: If TRUE, return a lower-triangular correlation table with one row per variable, like Stata's display. The default is a long table.
.wt: Optional weights: a numeric column of .data (or a vector with one value per row), like Stata's pwcorr [fweight/aweight]. Pearson correlations only. Rows with a missing weight are excluded with a message.
.wt_type: "frequency" (default) reports the sum of weights as n; "analytic" reports the number of complete observations with positive weights. Both give the same weighted correlation. Zero-weight rows are excluded. Like pwcorr(), fractional frequency weights are accepted, extending Stata's integer frequency-weight convention.

Value

An ungrouped tibble with one row per distinct variable pair and group: any grouping keys, var1, var2, r (Pearson correlation), and n (complete observations or summed frequency weight). Every pair within a group uses the same sample and has the same n. P-values are omitted. With .wide = TRUE, columns are grouping keys, variable, and one correlation column per variable; the diagonal is 1 and the upper triangle is NA. Undefined correlations are NA.

Details

Select at least two numeric or logical vector columns. When ... is omitted, nonnumeric columns and grouping keys are skipped, as in pwcorr(). Labelled numeric columns use their codes; tagged and SPSS user-missing values are excluded before determining the common sample. Missing values in unselected columns do not affect the sample.

Computed columns are evaluated on ungrouped working data before complete cases and groups are processed. Create columns with grouped dplyr::mutate() first for calculations that need to run separately within each group. Groups with fewer than two complete observations retain their pairs with correlations of NA; empty grouped inputs return zero rows. The input is not modified.

Like Stata's correlation mode, this function computes Pearson correlations. Stata's covariance, means, and display-formatting options are not implemented. Use stats::cov() for covariance and summ() for summaries. The tidy long table is an R interface choice; Stata prints a matrix.

References

Stata correlate manual.

Examples

corr(lifeexp, lexp, gnppc, safewater)
#> # A tibble: 3 x 4
#>   var1  var2          r     n
#>   <chr> <chr>     <dbl> <int>
#> 1 lexp  gnppc     0.777    53
#> 2 lexp  safewater 0.761    53
#> 3 gnppc safewater 0.657    53

corr(lifeexp, lexp, gnppc, safewater, .wide = TRUE)
#> # A tibble: 3 x 4
#>   variable   lexp  gnppc safewater
#>   <chr>     <dbl>  <dbl>     <dbl>
#> 1 lexp      1     NA            NA
#> 2 gnppc     0.777  1            NA
#> 3 safewater 0.761  0.657         1

corr(lifeexp, lexp, gnppc, safewater, .by = region)
#> # A tibble: 9 x 5
#>   region               var1  var2          r     n
#>   <dbl+lbl>            <chr> <chr>     <dbl> <int>
#> 1 1 [Europe & C. Asia] lexp  gnppc     0.796    41
#> 2 1 [Europe & C. Asia] lexp  safewater 0.789    41
#> 3 1 [Europe & C. Asia] gnppc safewater 0.653    41
#> 4 2 [North America]    lexp  gnppc     0.873     6
#> 5 2 [North America]    lexp  safewater 0.788     6
#> 6 2 [North America]    gnppc safewater 0.931     6
#> 7 3 [South America]    lexp  gnppc     0.484     6
#> 8 3 [South America]    lexp  safewater 0.713     6
#> 9 3 [South America]    gnppc safewater 0.897     6

corr(lifeexp, lexp, gnppc_thousands = gnppc / 1000)
#> # A tibble: 1 x 4
#>   var1  var2                r     n
#>   <chr> <chr>           <dbl> <int>
#> 1 lexp  gnppc_thousands 0.733    67

# Compare listwise and pairwise samples
survey <- tibble::tibble(
  x = c(1, 2, 3, 4), y = c(2, 3, NA, 5), z = c(1, NA, 4, 3)
)
corr(survey)
#> # A tibble: 3 x 4
#>   var1  var2      r     n
#>   <chr> <chr> <dbl> <int>
#> 1 x     y         1     2
#> 2 x     z         1     2
#> 3 y     z         1     2

pwcorr(survey)
#> # A tibble: 3 x 5
#>   var1  var2      r     n      p
#>   <chr> <chr> <dbl> <int>  <dbl>
#> 1 x     y     1         3  0    
#> 2 x     z     0.786     3  0.425
#> 3 y     z     1         2 NA