median(x, na.rm = F, weights = NULL)
bdNumeric.
Missing values (NA) are allowed.
x for computing weighted
median (see below). Does not have to be normalized to sum to 1.
If argument
weights is provided,
x is sorted and
weights are normalized to sum to 1. The median is then
the value of
x corresponding to the first cum-summed weight greater
than or equal to .5. In case a cum-summed weight is equal to .5, the
median is the average of the value of
x corresponding to that weight
and the value of
x corresponding to the first cumsumed weight
strictly greater than .5 (there may be intervening zero-valued
weights, which are ignored).
If
x contains any
NAs, the result will be
NA unless
na.rm=TRUE.
For long vectors or
bdVectors, using
quantile(x, .5) rather than
median(x) is significantly faster, because
quantile uses a partial sort, while
median uses a sort.
Note that
quantile(x, .5, weights = weights) gives a different
answer than
median(x, weights = weights).
Also,
quantile uses single precision.
algebra <- testscores[,3] # create sample object of algebra testscores
median(algebra) # median of scores
apply(testscores, 2, median) # vector of the medians of the columns
# in testscores data set