case.match {caseMatch}

Uses matching methods to select cases for qualitative analysis

Description

Uses matching methods to select cases for qualitative analysis

Usage

case.match(data, id.var, case.N = 2, distance = "mahalanobis", 
    design.type = "most similar", match.case = NULL, 
    greedy.match="pareto", number.of.matches.to.return = 1, 
    treatment.var = NULL, outcome.var= NULL, leaveout.vars = NULL, 
    max.variance = FALSE,  max.variance.outcome=FALSE,
    variance.tolerance = 0.1, max.spread = FALSE, 
    max.spread.outcome=FALSE, varweights = NULL)

Arguments

data

A data frame.

id.var

A string variable that uniquely identifies cases within the data

case.N

The number of cases to choose. Must be 1 or more.

distance

The distance metric, specified as a string. Options are "mahalanobis", "euclidean", or "standardized", where "standardized" means that variables are standardized by their standard deviations.

design.type

Should the algorithm pick cases that are most similar or most different? Specify either "most similar" or "most different" as a string.

match.case

If specified, this is the value of id.var of a specific case to match.

number.of.matches.to.return

How many matches to return.

greedy.match

Specifies which matches to return. Options are "pareto", "greedy", and "all". "all" keeps all matches. "pareto" matches eliminate 'redundant' matches where both units have better available matches. "greedy" matches keeps only the top matches in the data, but does eliminates best matches for some units since it uses a without replacement algorithm.

treatment.var

The name of the treatment variable, specified as a string.

outcome.var

The name of the outcome variable, specified as a string.

leaveout.vars

A vector of variables to not include in the matching.

max.variance

Should the cases be selected to maximize variance on treatment.var? If cases should be in opposite treatment conditions, specify max.variance=TRUE.

max.variance.outcome

Should the cases be selected to maximize variance on outcome.var? If cases should have opposite outcomes, specify max.variance.outcome=TRUE.

variance.tolerance

The proportion of cases to consider if max.variance is specified but there are too few cases that maximize the variance of treatment.var.

max.spread

Should the cases be selected to maximize "spread" on the treatment variable, meaning that cases are selected to be have evenly values from the min of treatment.var to the max?

max.spread.outcome

Should the cases be selected to maximize "spread" on the outcome variable, meaning that cases are selected to be have evenly values from the min of outcome.var to the max?

varweights

An optional vector of variable weights. It must line up with the columns of the data after id.var and leaveout.vars are removed. Optionally, element names can be included for varweights — if so, the function checks that the names are identical to (and line up with) the names of the matching variables. It will throw an error if they do not.

Details

case.match uses statistical matching to select cases in a quantitative data set for subsequent qualitative analysis in "most similar" and "most different" research designs.

Value

case.match returns a named list with the following elements:

cases

A table of the matched cases.

case.distances

A list of the distances between matched cases.

Author(s)

Rich Nielsen

References

Nielsen, Richard. 2016. "Case Selection via Matching," Sociological Methods and Research, 45 (3): 569-597. http://www.mit.edu/~rnielsen/Case

Examples

data(EU)
mvars <- c("socialist","rgdpc","FHc","FHp","trade")
dropvars <- c("countryname","population")

## In this example, I subset to the first 40 obs. to cut run-time
out <- case.match(data=EU[1:40,], id.var="countryname", leaveout.vars=dropvars,
             distance="mahalanobis", case.N=2, 
             number.of.matches.to.return=10,
             treatment.var="eu", max.variance=TRUE)
out$cases

## Not run: 
## All cases:
## Find the best matches of EU to non-EU countries
out <- case.match(data=EU, id.var="countryname", leaveout.vars=dropvars,
             distance="mahalanobis", case.N=2, 
             number.of.matches.to.return=10,
             treatment.var="eu", max.variance=TRUE)
out$cases

## Find the best matches while downweighting political variables
myvarweights <- c(1,1,.1,.1,.1)
names(myvarweights) <- c("rgdpc","trade","FHp","FHc","socialist")
myvarweights
(case.match(data=EU, id.var="countryname", leaveout.vars=dropvars,
             distance="mahalanobis", case.N=2, 
             number.of.matches.to.return=10, treatment.var="eu",
             max.variance=TRUE,varweights=myvarweights))$cases

## Find the best non-EU matches for Germany
tabGer <- case.match(data=EU, match.case="German Federal Republic", 
             id.var="countryname",leaveout.vars=dropvars,
             distance="mahalanobis", case.N=2, 
             number.of.matches.to.return=10,max.variance=TRUE,
             treatment.var="eu")

## End(Not run)