Generate Multi-row Tables Comparing Means/Medians/Frequencies of Multiple Variables Across Levels of One Categorical Variable (Survey Data)

Share:

Description

This function basically provides an alternative to making multiple calls to tabmeans.svy, tabmedians.svy, and tabfreq.svy, then using rbind to combine the results into a single table. Similar to tabmulti, but for survey data. Relies heavily on the 'survey' package [1,2].

Usage

1
2
3
4
5
6
7
8
9
tabmulti.svy(svy, xvarname, yvarnames, ymeasures = NULL, listwise.deletion = FALSE, 
             latex = FALSE, xlevels = NULL, ynames = yvarnames, ylevels = NULL, 
             mean.tests = "Wald", median.tests = "wilcoxon", freq.tests = "F", 
             decimals = 1, p.include = TRUE, p.decimals = c(2, 3), p.cuts = 0.01, 
             p.lowerbound = 0.001, p.leading0 = TRUE, p.avoid1 = FALSE, n.column = FALSE, 
             n.headings = TRUE, se = FALSE, compress = FALSE, parenth = "iqr", 
             text.label = NULL, parenth.sep = "-", bold.colnames = TRUE, 
             bold.varnames = FALSE, bold.varlevels = FALSE, variable.colname = "Variable", 
             print.html = FALSE, html.filename = "table1.html")

Arguments

svy

Survey design object created by a call to svydesign [1,2].

xvarname

Character string with name of column variable. Should be one of colnames(dataset).

yvarnames

Character string or vector of character strings with names of row variables. Each element should be one of colnames(dataset).

ymeasures

Character string or vector of character strings indicating whether each row variable should be summarized by mean, median, or frequency. For example, if yvarnames has length three and you wish to display frequencies for the first variable, means for the second, and medians for the third, you would set ymeasures to c("freq", "mean", "median"). If unspecified, function displays frequencies for any factor variable or numeric variable with five or fewer unique values, and means for numeric variables with more than five levels.

listwise.deletion

If TRUE, observations with missing values for any row variable are excluded entirely; if FALSE, all available data is used for each comparison. If FALSE, recommend also setting n to TRUE so table shows effective sample size for each comparison.

latex

If TRUE, object returned is formatted for printing in LaTeX using xtable [3]; if FALSE, formatted for copy-and-pasting from RStudio into a word processor.

xlevels

Optional character vector to label the levels of x. If unspecified, the function uses the values that x takes on.

ynames

Optional labels for the row variables.

ylevels

Character vector or list of character vectors to label the levels of the categorical row variables.

mean.tests

Character string or vector of character strings indicating what statistical tests should be used to compare means for each row variable for which a comparison of means is requested. Elements should be 'Wald' for Wald test or 'LRT' for likelihood ratio test.

median.tests

Character string or vector of character strings indicating what statistical tests should be used to compare medians for each row variable for which a comparison of medians is requested. Elements should be possible values for the 'test' input of the svyranktest function in the survey package [1,2]: 'wilcoxon' for Mann-Whitney U/Wilcoxon test of whether one group is from distribution that is stochastically greater than the other; 'vanderWaerden' for Van der Waerden test of whether the population distribution functions are equal; 'median' for Mood's test for whether the population medians are equal; and 'KruskalWallis' for Kruskal-Wallis test which is Mann-Whitney U/Wilcoxon generalized to three or more groups.

freq.tests

Character string or vector of character strings indicating what statistical tests should be used to compare distributions of each categorical row variable across levels of the column variable. Elements should be possible values for the 'statistic' input of the svychisq function in the survey package [1,2]: 'F', 'Chisq', 'Wald', 'adjWald', 'lincom', or 'saddlepoint'.

decimals

Number of decimal places for various cell entries, such as means and percentages. Does not affect p-values.

p.include

If FALSE, statistical test is not performed and p-value is not returned.

p.decimals

Number of decimal places for p-values. If a vector is provided rather than a single value, number of decimal places will depend on what range the p-value lies in. See p.cuts.

p.cuts

Cut-point(s) to control number of decimal places used for p-values. For example, by default p.cuts is 0.1 and p.decimals is c(2, 3). This means that p-values in the range [0.1, 1] will be printed to two decimal places, while p-values in the range [0, 0.1) will be printed to three decimal places.

p.lowerbound

Controls cut-point at which p-values are no longer printed as their value, but rather <lowerbound. For example, by default p.lowerbound is 0.001. Under this setting, p-values less than 0.001 are printed as <0.001.

p.leading0

If TRUE, p-values are printed with 0 before decimal place; if FALSE, the leading 0 is omitted.

p.avoid1

If TRUE, p-values rounded to 1 are not printed as 1, but as >0.99 (or similarly depending on values for p.decimals and p.cuts).

n.column

If TRUE, the table will have a column for (unweighted) sample size.

n.headings

If TRUE, the table will indicate the (unweighted) sample size overall and in each group in parentheses after the column headings.

se

If TRUE, the table will present mean (standard error) rather than mean (standard deviation) for continuous row variables.

compress

If TRUE, categorical row variables with two levels will have a single row for n (percent) for the higher level. For example, if a row variable is sex, with 0 for females and 1 for males, setting compress = TRUE would result in the sex row showing n (percent) for males only. If FALSE, the table would have two rows for sex, one showing n (percent) for males and another sohwing n (percent) for females.

parenth

For median comparisons, controls what values (if any) are placed in parentheses after the medians in each cell. Possible choices are as follows: 'minmax' for minimum and maximum; 'range' for difference between minimum and maximum; 'q1q3' for first and third quartiles; 'iqr' for difference between first and third quartiles; or 'none' for no parentheses at all.

text.label

For median comparisons, optional text to put after the variable name. For example, if parenth is 'q1q3' and yname is 'BMI' the default label would be 'BMI, Median (Q1-Q3)'. You might prefer to set text.label to something like 'Med (Quartile 1-Quartile 3)' instead.

parenth.sep

For median comparisons, optional character specifying the separator for the two numbers in parentheses when parenth is set to 'minmax' or 'q1q3'. The default is a dash, so values in the table are formatted as Median (Lower-Upper). If you set parenth.sep to ', ' the values in the table would instead be formatted as Median (Lower, Upper).

bold.colnames

If TRUE, column headings are printed in bold font. Only applies if latex = TRUE.

bold.varnames

If TRUE, variable name in the first column of the table is printed in bold font. Only applies if latex = TRUE.

bold.varlevels

If TRUE, levels of categorical y variables are printed in bold font. Only applies if latex = TRUE.

variable.colname

Character string with desired heading for first column of table, which shows the y variable name and levels.

print.html

If TRUE, function prints a .html file to the current working directory.

html.filename

Character string indicating the name of the .html file that gets printed if print.html is set to TRUE.

Details

Please see help files for tabmeans.svy, tabmedians.svy, and tabfreq.svy for details on statistical tests.

Value

A character matrix comparing means/medians/frequencies of row variables across levels of the column variable. If latex is set to TRUE, the character matrix will be formatted for inserting into a Markdown/Sweave/knitr report using the xtable package [3].

Note

If you wish to paste your tables into Word, you can use either of these approaches:

1. Use the write.cb function in the Kmisc package [4]. If your table is stored in a character matrix named table1, use write.cb(table1) to copy the table to your clipboard. Paste the result into Word, then highlight the text and go to Insert - Table - Convert Text to Table... OK.

2. Set the print.html input to TRUE. This will result in a .html file writing to your current working directory. When you open this file, you will see a nice looking table that you can copy and paste into Word. You can control the name of this file with the html.filename input.

If you wish to use LaTeX, R Markdown, knitr, Sweave, etc., please see the package vignette for examples. In most cases, you have to set the latex input to TRUE and then use the xtable package [3].

If you have suggestions for additional options or features, or if you would like some help using any function in the package tab, please e-mail me at vandomed@gmail.com. Thanks!

Author(s)

Dane R. Van Domelen

References

1. Lumley T (2012). survey: analysis of complex survey samples. R package version 3.28-2, https://cran.r-project.org/package=survey.

2. Lumley T (2014). Analysis of complex survey samples. Journal of Statistical Software 9(1): 1-19.

3. Dahl DB (2013). xtable: Export tables to LaTeX or HTML. R package version 1.7-1, https://cran.r-project.org/package=xtable.

4. Kevin Ushey (2013). Kmisc: Kevin Miscellaneous. R package version 0.5.0. https://CRAN.R-project.org/package=Kmisc.

Acknowledgment: This material is based upon work supported by the National Science Foundation Graduate Research Fellowship under Grant No. DGE-0940903.

See Also

svydesign, svyglm, svychisq, svyquantile, svyranktest, tabfreq, tabmeans, tabmedians, tabglm, tabcox, tabgee, tabfreq.svy, tabmeans.svy, tabmedians.svy, tabglm.svy

Examples

1