The numeric variable.
Categories of the ordinal variable, the same length.
ML and stdErr (see PolycorOptions).
Polyserial correlation of a numeric variable and an ordinal one. By default the two-step (ad hoc) estimate.
Equivalent to polycor::polyserial(x, y) in R.
Pairs with a missing value are dropped. The categories of y are its sorted distinct values, coded 1, 2, ...;
with fewer than two, rho is NaN (and the result has no other fields). A two-step estimate beyond ±0.9999 is set
to ±0.9999 with a warning.
With ML, the correlation and the thresholds are estimated together by maximum likelihood; with stdErr, the
result also has their variances, polycor's chi-square test of bivariate normality and the thresholds (with stdErr
alone, the correlation is maximised given the two-step thresholds). The exact maximum is returned (R's optimisers
stop about 1e-4 short of it). If the likelihood increases up to |rho| = 0.9999, rho is ±0.9999 with a warning.
The numeric variable.
Categories of the ordinal variable, the same length.
Optionaloptions: PolycorOptions
ML and stdErr (see PolycorOptions).
Polyserial correlation of a numeric variable and an ordinal one. By default the two-step (ad hoc) estimate. Equivalent to
polycor::polyserial(x, y)in R.Pairs with a missing value are dropped. The categories of
yare its sorted distinct values, coded 1, 2, ...; with fewer than two,rhoisNaN(and the result has no other fields). A two-step estimate beyond ±0.9999 is set to ±0.9999 with a warning.With
ML, the correlation and the thresholds are estimated together by maximum likelihood; withstdErr, the result also has their variances, polycor's chi-square test of bivariate normality and the thresholds (withstdErralone, the correlation is maximised given the two-step thresholds). The exact maximum is returned (R's optimisers stop about 1e-4 short of it). If the likelihood increases up to |rho| = 0.9999,rhois ±0.9999 with a warning.