| Title: | Correspondence Analysis Plot and Associations Visualisation |
|---|---|
| Description: | Performs a Correspondence Analysis (CA) on a contingency table and creates a scatterplot of the row and column points on the selected dimensions. Optionally, the function can add segments to the plot to visualize significant associations between row and column categories on the basis of positive (unadjusted) standardized residuals larger than a given threshold. |
| Authors: | Gianmarco Alberti [aut, cre] |
| Maintainer: | Gianmarco Alberti <[email protected]> |
| License: | GPL (>= 2) |
| Version: | 0.1 |
| Built: | 2024-11-02 03:11:37 UTC |
| Source: | https://github.com/cran/caresid |
Performs a Correspondence Analysis (CA) on a contingency table and creates a scatterplot
of the row and column points on the selected dimensions. Optionally, the function
can add segments to the plot to visualize significant associations between row and column
categories on the basis of positive (unadjusted) standardized residuals larger than a given threshold. The
segments can be optionally labelled with the corresponding residual value.
Visit this LINK to access the package's vignette.
caresid( cross.tab, dim1 = 1, dim2 = 2, segments = FALSE, category = NULL, mult.comp = FALSE, label.residuals = FALSE, residual.label.size = 2, dot.size = 1, dot.label.size = 2.5, axis.label.size = 9, square = FALSE )caresid( cross.tab, dim1 = 1, dim2 = 2, segments = FALSE, category = NULL, mult.comp = FALSE, label.residuals = FALSE, residual.label.size = 2, dot.size = 1, dot.label.size = 2.5, axis.label.size = 9, square = FALSE )
cross.tab |
A dataframe representing the input contingency table. |
dim1 |
The first dimension to plot (default is 1). |
dim2 |
The second dimension to plot (default is 2). |
segments |
Logical. If TRUE, add segments to the plot to connect row to column points (or viceversa) with positive standardized residuals larger than a given threshold (default is FALSE). |
category |
Character vector. If provided, only add segments from that/those row (or column) category(ies) to the column (or row) categories where the corresponding standardised residuals are positive and larger than a given threshold. If NULL (default) all the categories are considered. |
mult.comp |
Logical. If TRUE, adjust the residuals' significance threshold for multiple comparisons using Sidak's method (default is FALSE). |
label.residuals |
Logical. If TRUE, the value of the positive standardised residual will be shown as a label at the midpoint of every segment (default is FALSE). |
residual.label.size |
Numeric. The size of the residuals' label (default is 2). |
dot.size |
Numeric. The size of the scatterplot's points (default is 1). |
dot.label.size |
Numeric. The size of the points' label (default is 2.5). |
axis.label.size |
Numeric. The size of the axis labels (default is 9). |
square |
Logical. If TRUE, set the ratio of y to x to 1 (default is FALSE). |
If the segment argument is FALSE (default), a regular symmetric CA biplot is rendered.
If the segment argument is TRUE, the function adds segments to the plot to connect
row and column points with positive (unadjusted) standardized residuals larger than a given threshold, indicating
a significant association. The threshold is 1.96 if mult.comp is FALSE, and is
adjusted for multiple comparisons if mult.comp is TRUE.
In the latter case, the threshold for significant residuals is calculated using the Sidak's method.
It is based on an adjusted 0.05 alpha level which is calculated as 1-(1 - 0.05)^(1/(nr*nc)),
where nr and nc are the number of rows and columns in the table respectively.
The adjusted alpha is then converted to a critical two-tailed z value (see Beasley-Schumacker 1995).
Please note, all the visualised associations (if any) are significant at least at alpha 0.05.
Optionally, the residual segments can be labelled with the corresponding residual value by setting
the label.residuals to TRUE.
The idea of connecting points in a CA plot based on the value of standardized residuals can serve to visually highlight certain associations in your data. However, please note that while this function can help visualize the associations in the contingency table, it does not replace other formal approaches for the interpretation of the CA scatterplot and formal statistical tests for assessing the significance and strength of the association.
A list with two elements:
stand.residuals contains the unadjusted standardized residuals for all cells.
resid.sign.thres contains the threshold used to determine significant residuals.
Beasley TM and Schumacker RE (1995), Multiple Regression Approach to Analyzing Contingency Tables: Post Hoc and Planned Comparison Procedures, The Journal of Experimental Education, 64(1): 86, 89.
# Create a toy dataset (famous Eye-color Hair-color dataset) mytable <- structure(list(BLACK_H = c(68, 20, 15, 5), BROWN_H = c(119, 84, 54, 29), RED_H = c(26, 17, 14, 14), BLOND_H = c(7, 94, 10, 16)), class = "data.frame", row.names = c("Brown_E", "Blue_E", "Hazel_E", "Green_E")) # EXAMPLE 1 # Run the function: result <- caresid(mytable, segments=TRUE) # EXAMPLE 2 # As above, but adjusting for multiple comparisons: result <- caresid(mytable, segments=TRUE, mult.comp=TRUE) # EXAMPLE 3 # As in the first example, but selecting only 2 row categories; # residual labels are shown: result <- caresid(mytable, segments=TRUE, category=c("Brown_E", "Green_E"), label.residuals=TRUE)# Create a toy dataset (famous Eye-color Hair-color dataset) mytable <- structure(list(BLACK_H = c(68, 20, 15, 5), BROWN_H = c(119, 84, 54, 29), RED_H = c(26, 17, 14, 14), BLOND_H = c(7, 94, 10, 16)), class = "data.frame", row.names = c("Brown_E", "Blue_E", "Hazel_E", "Green_E")) # EXAMPLE 1 # Run the function: result <- caresid(mytable, segments=TRUE) # EXAMPLE 2 # As above, but adjusting for multiple comparisons: result <- caresid(mytable, segments=TRUE, mult.comp=TRUE) # EXAMPLE 3 # As in the first example, but selecting only 2 row categories; # residual labels are shown: result <- caresid(mytable, segments=TRUE, category=c("Brown_E", "Green_E"), label.residuals=TRUE)