The acs package provides a new S4 method for
standard division operations using "/" notation. However, due to the
nature of estimates and errors, there are actually two types of
division, with slightly different meanings: depending on which
variables are being divided, the process may be either a
"proportion"-type division (in which the numerator is a subset of the
denominator) or a "ratio"-type division (in which this is not the
case). When dividing with standard "a/b" notation, the package will
always use the more conservative ratio-type procedure.
When appropriate, "proportion"-type division may be desirable, as it
results in lower standarard errors. To allow users to specify which
type of division to use for acs objects, the package includes a new
"divide.acs" function. (See details.)
divide.acs(numerator, denominator, method="ratio", verbose=T, output="result")an acs object to divide
an acs object to divide by
either "ratio" (the default) or "proportion", to indicate what kind of division is desired
whether to provide additional warnings or just shut up
either "result" (the default), "div.method", or "both"
Returns a new acs object with the results of the division (the default), or (when result="div.method") a martix with diagnostic information, or (when result="both"), a list with both of these objects (the first name $result and the second $div.method).
In certain cases, "proportion-style" division will fail, due to the creation of a negative number under a square root when calculating new standard errors. To address this problem and prevent unnecessary NaN values in the standard.errors, the package implements the recommended Census practice of simply using "ratio-style" division in those cases.
If method="proportion" (not the default) and verbose=T (the default),
division.acs will provide a warning to indicate when
"ratio-style" division has been used, including the number of standard
error cells so affected. Users wishing to examine a detailed,
cell-by-cell report may run divide.acs with the
output="div.method" of output="both" to get additional diagnostic
information.
See "A Compass for Understanding and Using American Community Survey Data" below for details on when this substitution is recommended.
"A Compass for Understanding and Using American Community Survey Data: What State and Local Governments Need to Know." Washington, DC: U.S. Census Bureau. 2009. http://www.census.gov/library/publications/2009/acs/state-and-local.html.