selectMax {cnaOpt}

Select the con-cov optima from a 'conCovOpt' object that maximize a specified optimality criterion

Description

selectMax selects the optima from a 'conCovOpt' object that maximize a specified optimality criterion (cf. Baumgartner and Ambuehl 2021).

Usage

selectMax(x, crit = quote(con * cov), cond = quote(TRUE), warn = TRUE)
multipleMax(x, outcome)

Arguments

x

An object output by conCovOpt.

crit

Quoted expression specifying a numeric criterion to be maximized when selecting from the con-cov optima that meet criterion cond, for example, min(con,cov) or 0.8*con + 0.2*cov, etc.

cond

Quoted expression specifying a logical criterion to be imposed on the con-cov optima in x before selecting the optima maximizing crit, for example, con > 0.85 or con > cov, etc.

warn

Logical; if TRUE, selectMax() returns a warning if no solution is found.

outcome

A character string specifying a single outcome value in the original data.

Details

While conCovOpt identifies all con-cov optima in an analyzed data set, selectMax selects those optima from a 'conCovOpt' object x that comply with a logical condition cond and fare best according to the numeric optimality criterion crit. The default is to select so-called con-cov maxima, meaning con-cov optima with highest product of consistency and coverage. But the argument crit allows for specifying any other numeric optimality criterion, e.g. min(con, cov), max(con, cov), or 0.8*con + 0.2*cov, etc. (see Baumgartner and Ambuehl 2021). If x contains multiple outcomes, the selection of the best con-cov optima is done separately for each outcome.

As of package version 0.5.0, the function multipleMax is obsolete. It is kept for backwards compatibility only.

Via the column id in the output of selectMax it is possible to select one among many equally good maxima, for instance, by means of reprodAssign (see the examples below).

Value

selectMax returns an object of class 'selectMax'.

References

Baumgartner, Michael and Mathias Ambuehl. 2021. “Optimizing Consistency and Coverage in Configurational Causal Modeling.” Sociological Methods & Research.
doi:10.1177/0049124121995554.

See Also

conCovOpt, reprodAssign

See also examples in conCovOpt.

Examples

dat1 <- d.autonomy[15:30, c("EM","SP","CO","AU")]
(cco1 <- conCovOpt(dat1, outcome = "AU"))
selectMax(cco1)
selectMax(cco1, cond = quote(con > 0.95))
selectMax(cco1, cond = quote(cov > 0.98))
selectMax(cco1, crit = quote(min(con, cov)))
selectMax(cco1, crit = quote(max(con, cov)), cond = quote(cov > 0.9))

# Multiple equally good maxima.
(cco2 <- conCovOpt(dat1, outcome = "AU")) 
(sm2 <- selectMax(cco2, cond = quote(con > 0.93)))
# Each maximum corresponds to a different rep-assignment, which can be selected
# using the id argument.
reprodAssign(sm2, "AU", id = 10)
reprodAssign(sm2, "AU", id = 11)
reprodAssign(sm2, "AU", id = 13)