Construct a matched dataset from a matchit object
match.data and get_matches create a data frame with additional variables for the distance measure, matching weights, and subclasses after matching. This dataset can be used to estimate treatment effects after matching or subclassification. get_matches is most useful after matching with replacement; otherwise, match.data is more flexible. See Details below for the difference between them.
match.data(object, group = "all", distance = "distance",
weights = "weights", subclass = "subclass",
data = NULL, include.s.weights = TRUE,
drop.unmatched = TRUE)
get_matches(object, distance = "distance",
weights = "weights", subclass = "subclass",
id = "id", data = NULL, include.s.weights = TRUE)object |
a |
group |
which group should comprise the matched dataset: |
distance |
a string containing the name that should be given to the variable containing the distance measure in the data frame output. Default is |
weights |
a string containing the name that should be given to the variable containing the matching weights in the data frame output. Default is |
subclass |
a string containing the name that should be given to the variable containing the subclasses or matched pair membership in the data frame output. Default is |
id |
a string containing the name that should be given to the variable containing the unit IDs in the data frame output. Default is |
data |
a data frame containing the original dataset to which the computed output variables ( |
include.s.weights |
|
drop.unmatched |
|
match.data creates a dataset with one row per unit. It will be identical to the dataset supplied except that several new columns will be added containing information related to the matching. When drop.unmatched = TRUE, the default, units with weights of zero, which are those units that were discarded by common support or the caliper or were simply not matched, will be dropped from the dataset, leaving only the subset of matched units. The idea is for the output of match.data to be used as the dataset input in calls to glm or similar to estimate treatment effects in the matched sample. It is important to include the weights in the estimation of the effect and its standard error. The subclass column, when created, contains par or subclass membership and should be used to estimate the effect and its standard error. Subclasses will only be included if there is a subclass component in the matchit object, which does not occur with matching with replacement, in which case get_matches should be used. See vignette("estimating-effects") for information on how to use match.data output to estimate effects.
get_matches is similar to match.data; the primary difference occurs when matching is performed with replacement, i.e., when units do not belong to a single matched pair. In this case, the output of get_matches will be a dataset that contains one row per unit for each pair they are a part of. For example, if matching was performed with replacement and a control unit was matched to two treated units, that control unit will have two rows in the output dataset, one for each pair it is a part of. Weights are computed for each row, and are equal to the inverse of the number of control units in each control unit's subclass. Unmatched units are dropped. An additional column with unit IDs will be created (named using the id argument) to identify when the same unit is present in multiple rows. This dataset structure allows for the inclusion of both subclass membership and repeated use of units, unlike the output of match.data, which lacks subclass membership when matching is done with replacement. A match.matrix component of the matchit object must be present to use get_matches; in some forms of matching, it is absent, in which case match.data should be used instead. See vignette("estimating-effects") for information on how to use get_matches output to estimate effects after matching with replacement.
A data frame containing the data supplied in the data argument or in the original call to matchit with the computed output variables appended as additional columns, named according the arguments above. For match.data, the group and drop.unmatched arguments control whether only subsets of the data are returned. See Details above for how match.data and get_matches differ. Note that get_matches sorts the data by subclass and treatment status, unlike match.data, which uses the order of the data.
The returned data frame will contain the variables in the original data set or dataset supplied to data, and the following columns:
distance |
The propensity score, if estimated or supplied to the |
weights |
The computed matching weights. These must be used in effect estimation to correctly incorporate the matching. |
subclass |
Matching strata membership. Units with the same value are in the same stratum. |
id |
The ID of each unit, corresponding to the row names in the original data or dataset supplied to |
These columns will take on the name supplied to the corresponding arguments in the call to match.data or get_matches. See Examples for an example of rename the distance column to "prop.score".
If data or the original dataset supplied to matchit was a data.table or tbl, the match.data output will have the same class, but the get_matches output will always be a base R data.frame.
vignette("estimating-effects") for uses of match.data() and get_matches() in estimating treatment effects.
data("lalonde")
# 4:1 matching w/replacement
m.out1 <- matchit(treat ~ age + educ + married +
race + nodegree + re74 + re75,
data = lalonde, replace = TRUE,
caliper = .05, ratio = 4)
m.data1 <- match.data(m.out1, data = lalonde,
distance = "prop.score")
dim(m.data1) #one row per matched unit
head(m.data1, 10)
g.matches1 <- get_matches(m.out1, data = lalonde,
distance = "prop.score")
dim(g.matches1) #multiple rows per matched unit
head(g.matches1, 10)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.