MDmiss: Mahalanobis distance (MD) for data with missing values
Description
For each observation the missing dimensions are omitted
before calculating the MD. The MD contains a correction
factor \(p/q\) to account for the number of observed values,
where \(p\) is the number of variables and \(q\) is the number of
observed dimensions for the particular observation.
Usage
MDmiss(data, center, cov)
Value
The function returns a vector of the (squared) Mahalanobis distances.
Arguments
data
the data as a dataframe or matrix.
center
the center to be used (may not contain missing values).
cov
the covariance to be used (may not contain missing values).
Author
Beat Hulliger
Details
The function loops over the observations. This is not optimal if
only a few missingness patterns occur. If no missing values occur
the function returns the Mahalanobis distance.
References
Béguin, C., and Hulliger, B. (2004). Multivariate outlier detection
in incomplete survey data: The epidemic algorithm and transformed rank correlations.
Journal of the Royal Statistical Society, A167 (Part 2.), pp. 275-294.