sieve function

Extended Sieve Plots

(Extended) sieve displays for n-way contingency tables: plots rectangles with areas proportional to the expected cell frequencies and filled with a number of squares equal to the observed frequencies. Thus, the densities visualize the deviations of the observed from the expected values.

UTF-8

## Default S3 method:
sieve(x, condvars = NULL, gp = NULL, shade = NULL,
  legend = FALSE, split_vertical = NULL, direction = NULL, spacing = NULL,
  spacing_args = list(), sievetype = c("observed","expected"),
  gp_tile = gpar(), scale = 1, main = NULL, sub = NULL, ...)
## S3 method for class 'formula'
sieve(formula, data, ..., main = NULL, sub = NULL, subset = NULL)

Arguments

• x: a contingency table in array form, with optional category labels specified in the dimnames(x) attribute.

• condvars: vector of integers or character strings indicating conditioning variables, if any. The table will be permuted to order them first.

• formula: a formula specifying the variables used to create a contingency table from data. For convenience, conditioning formulas can be specified; the conditioning variables will then be used first for splitting. Formulas for sieve displays (unlike those for doubledecker plots) have no response variable.

• data: either a data frame, or an object of class "table"

  or "ftable".

• subset: an optional vector specifying a subset of observations to be used.

• shade: logical specifying whether gp should be used or not (see gp). If TRUE and expected is unspecified, a default model is fitted: if condvars is specified, a corresponding conditional independence model, and else the total independence model. If shade is NULL (default), gp is used if specified.

• sievetype: logical indicating whether rectangles should be filled according to observed or expected frequencies.

• gp: object of class "gpar", shading function or a corresponding generating function (see details of strucplot and shadings). Components of "gpar"

  objects are recycled as needed along the last splitting dimension. The default is a modified version of shading_Friendly: if sievetype is "observed", cells with positive residuals are painted with a red sieve, and cells with negative residuals with a blue one. If sievetype is "expected", the sieves' color is gray. Ignored if shade = FALSE.

• gp_tile: object of class "gpar", controlling the appearance of all static

  elements of the cells (e.g., border and fill color).

• scale: scaling factor for the sieve.

• legend: either a legend-generating function, a legend function (see details of strucplot and legends), or a logical value. If legend is NULL or TRUE and gp is a function, legend defaults to legend_resbased.

• split_vertical: vector of logicals of length $k$, where $k$

  is the number of margins of x (default: FALSE). Values are recycled as needed. A TRUE component indicates that the tile(s) of the corresponding dimension should be split vertically, FALSE

  means horizontal splits. Ignored if direction is not NULL.

• direction: character vector of length $k$, where $k$ is the number of margins of x (values are recycled as needed). For each component, a value of "h" indicates that the tile(s) of the corresponding dimension should be split horizontally, whereas "v" indicates vertical split(s).

• spacing: spacing object, spacing function, or corresponding generating function (see strucplot for more information). The default is no spacing at all if x has two dimensions, and spacing_increase for more dimensions.

• spacing_args: list of arguments for the generating function, if specified (see strucplot for more information).

• main, sub: either a logical, or a character string used for plotting the main (sub) title. If logical and TRUE, the name of the data object is used.

• ...: Other arguments passed to strucplot

Details

sieve is a generic function which currently has a default method and a formula interface. Both are high-level interfaces to the strucplot function, and produce (extended) sieve displays. Most of the functionality is described there, such as specification of the independence model, labeling, legend, spacing, shading, and other graphical parameters.

The layout is very flexible: the specification of shading, labeling, spacing, and legend is modularized (see strucplot for details).

Returns

The "structable" visualized is returned invisibly.

Note

To be faithful to the original definition by Riedwyl & Schüpbach, the default is to have no spacing between the tiles for two-way tables.

References

H. Riedwyl & M. Schüpbach (1994), Parquet diagram to plot contingency tables. In F. Faulbaum (ed.), Softstat '93: Advances in Statistical Software, 293--299. Gustav Fischer, New York.

M. Friendly (2000), Visualizing Categorical Data, SAS Institute, Cary, NC.

Meyer, D., Zeileis, A., and Hornik, K. (2006), The strucplot framework: Visualizing multi-way contingency tables with vcd. Journal of Statistical Software, 17(3) , 1-48. tools:::Rd_expr_doi("10.18637/jss.v017.i03") and available as vignette("strucplot").

Author(s)

David Meyer David.Meyer@R-project.org

See Also

assoc, strucplot, mosaic, structable, doubledecker

Examples

data("HairEyeColor")

## aggregate over 'sex':
(haireye <- margin.table(HairEyeColor, c(2,1)))

## plot expected values:
sieve(haireye, sievetype = "expected", shade = TRUE)

## plot observed table:
sieve(haireye, shade = TRUE)

## plot complete diagram:
sieve(HairEyeColor, shade = TRUE)

## example with observed values in the cells:
sieve(haireye, shade = TRUE, labeling = labeling_values,
               gp_text = gpar(fontface = 2))

## example with expected values in the cells:
sieve(haireye, shade = TRUE, labeling = labeling_values,
               value_type = "expected", gp_text = gpar(fontface = 2))

## an example for the formula interface:
data("VisualAcuity")
sieve(Freq ~ right + left,  data = VisualAcuity)