microaggregation {sdcMicro} | R Documentation |
Function to perform various methods of microaggregation.
microaggregation(x, method = "pca", aggr = 3, nc = 8, clustermethod = "clara", opt = FALSE, measure = "mean", trim = 0, varsort = 1, transf = "log", blow = TRUE, blowxm = 0)
x |
data frame or matrix |
method |
pca, onedims, single, simple, clustpca, pppca, clustpppca, mdav, clustmcdpca, influence, mcdpca |
aggr |
aggregation level (default=3) |
nc |
number of cluster, if the chosen method performs cluster analysis |
clustermethod |
clustermethod, if necessary |
opt |
|
measure |
aggregation statistic, mean, median, trim, onestep (default = mean) |
trim |
trimming percentage, if measure=trim |
varsort |
variable for sorting, if method= single |
transf |
transformation for data x |
blow |
if TRUE, the microaggregated data will have the same dimension as the original data set |
blowxm |
the microaggregated data with the same dimension as the original one. |
On http://neon.vb.cbs.nl/casc/Glossary.htm we can find the “official” definition of microaggregation:
Records are grouped based on a proximity measure of variables of interest, and the same small groups of records are used in calculating aggregates for those variables. The aggregates are released instead of the individual record values.
While for the proximity measure very different concepts can be used, microaggregation is naturally done with the mean. Nevertheless, other measures of location can be used for aggregation, especially when the group size for aggregation has been taken higher than 3. Since the median seems to be unsuitable for microaggregation due to it's rather high breakdown point, other mesures which are included can be chosen.
This function contains also a method with which the data can be clustered with a variety of different clustering algorithms. Clustering observations before applying microaggregation might be useful. Note, that the data are automatically log-transformed and standardised before clustering because most of the clustering algorithms performs better on log-transformed and standardised data.
The usage of clustering method ‘Mclust’ requires package mclust02, which must be loaded first. The package is not loaded automatically, since the package is not under GPL but on a differnt licence.
The are some projection methods for microaggregation included. The robust version ‘pppca’ or ‘clustpppca’ (clustering at first) are fast implementations and provide almost everytime the best results.
Univariate statistics are preserved best with the individual ranking method (we called them ‘onedims’), but multivariate statistics are strong affected.
With method simple one can apply microaggregation directly on the (unsorted) data and is useful for the comparison with other methods, i.e. reply the question how much better is a sorting of the data before aggregation.
If blow is set to FALSE, the result will be a data set with dimension n divided by aggr.
x |
original data |
method |
method |
clustering |
TRUE, if a clustering is done before microaggregation |
aggr |
aggregation level |
nc |
number of clusters, if a clustering method is chosen |
xm |
aggregated data set |
roundxm |
rounded aggregated data set (to integers) |
clustermethod |
clustermethod, if a cluster method is chosen |
measure |
proximity measure for aggregation |
trim |
trimming, if proximity measure ‘trim’ is chosen |
varsort |
information about the variable which is chosen when using method ‘single’ |
transf |
transformation used, when clustering is applied first |
blow |
TRUE, blowxm is calculated |
blowxm |
microaggregated data with the same dimension as the original data set |
fot |
correction factor, necessary if totals calculated and n divided by aggr is not an integer. |
Matthias Templ
http://www.springerlink.com/content/v257655u88w2/?sortorder=asc&p_o=20
summary.micro
, plotMicro
, valTable
data(Tarragona) m1 <- microaggregation(Tarragona, method="onedims", aggr=3) ## summary(m1)