ComparisonTable: Comparison Tables

Description

Produce a table of comparisons for reports and manuscripts.

Usage

cattable(data, vars, byVar, fisher=NULL,
   cmh=NULL, row.score=NULL, col.score=NULL,
   mcnemar=NULL, correct=NULL,
   odds=NULL, row.p=TRUE, alpha=0.05, minl=5)
   
conttable(data, vars, byVar,
   normal = NULL, var.equal = NULL,
   median = NULL,
   odds = NULL, odds.scale=NULL, odds.unit=NULL,
   alpha = 0.05, B=1000, seed=NULL)
   
catconttable(data, vars, byVar, vars.cat=NULL, fisher=NULL,
   cmh=NULL, row.score=NULL, col.score=NULL,
   normal = NULL, var.equal = NULL,
   median=NULL, odds=NULL, odds.scale=NULL, odds.unit=NULL,
   row.p=TRUE, alpha=0.05, B=1000, seed=NULL)

Arguments

data

A ccf.df or data.frame with the variables to be compared and the grouping variable.

vars

A character vector naming the variables to be compared

vars.cat

A character vector that can be used to specify which, if any, numeric variables in vars should be treated as categorical.

byVar

A character(1) giving the grouping variable. This allows more than one level. Numeric variables are coerced to factors.

fisher

A character vector giving the names of variables that should be compared with Fisher's Exact test. Currently, there is no implementation to determine this automatically.

cmh

A character vector giving the names of variables that should be compared with Manthel-Haenszel's Test for Linear Trend. This is not yet written and will be ignored.

row.score

Currently ignored

col.score

Currently ignored

mcnemar

a character vector giving the names of variables that should be compared using McNemar's test.

correct

Character vector giving the variables for which a continuity correction should be applied in McNemar's test.

odds

A character vector giving the names of variables for which odds ratios should be calculated. For categorical measures, this is the primary test of comparison. For numeric measures, this is calculated in addition to another test.

odds.scale

For numeric variables only. A list with named elements that gives the scale on which the odds ratio should be presented. For example, if the odds for variable x should be presented in 5 year increments, we would use odd

odds.unit

For numeric variables only. A list with named elements that gives the units on which the odds ratio should be presented. For example, if the odds of variable x should be presented in 5 year increments, we would use odd

row.p

Toggles if row or column proportions are calculated.

normal

A character vector that assigns variables in vars as normally distributed.

var.equal

A character vector that assigns variables in vars as having equal variance. This is used to determine the proper form of a t-test.

median

A character vector that assigns variables that shoudl be summarized with a median, quartiles, or min and max.

alpha

Significance levels for tests.

The number of Bootstrap samples for bootstrapped confidence intervals.

seed

The seed to use in starting the Bootstrapping.

minl

Minimum length for levels abbreviations. The function abbreviate is used to create unique rownames for each level of a variable in the output data frame. If the abbreviations are short, they may not be readable. This allow

Value

Returns an object of class ctable. The ctable class inherits the data.frame class and may be manipulated prior to printing to a report using write.ctable.

Details

catconttable is a wrapper that determines the type of variable and calls either cattable or conttable as appropriate. For this to work properly, all factor variables must be defined before the function call. In contrast, if cattable is called directly, variables are coerced to factors, which could lead to peculiar results if a numeric value is given.

Examples

Run this code

#Read in the delivery dataset from the CCFmisc library
data(Delivery)

#Use conttable to summarize maternal age, ga weeks, weight (grams) 
#and grava by delivery type.  The dataset name is specified under the "data="
#option, the variables of interest are listed under "vars=", and the K-level by variable 
#is specified under "byVar=".

#Default is to report mean and bootstrapped 95% CI for mean.  Tests of location are by 
#default either Wilcoxon rank sum (K=2) or Kruskal-Wallis (K>2) rank sum.  The "seed="
#option allows for reproducibility by setting the seed for getting bootstrapped samples.

d_type.contable <- conttable(data=Delivery,
                             vars=c("maternal.age", "ga.weeks", "wt.gram", "grava"),
                             byVar="delivery.type")

#Specifying weights by delivery type as a normally distributed variables, reports means, 
#standard deviations and a t-test of equality of the means for delivery type.  Variables listed 
#under "var.equal=" are assumed to have equal variances in all levels of byVar.  Otherwise, 
#variances are allowed to be unequal.

d_type.conttable <- conttable(data=Delivery,
                              vars=c("maternal.age", "ga.weeks", "wt.gram", "grava", "apgar1"),
                              byVar="delivery.type",
                              normal=c("wt.gram", "maternal.age"),
                              var.equal="ga.weeks")

#List variables under "median=" to report median, 25th and 75th percentiles.

d_type.conttable <- conttable(data=Delivery,
                              vars=c("maternal.age", "ga.weeks", "wt.gram", "grava", "apgar1"),
                              byVar="delivery.type",
                              normal=c("wt.gram", "maternal.age"),
                              var.equal="ga.weeks",
                              median=c("grava","apgar1"))

#Use cattable to summarize child sex, laceration, and laceration degree by delivery type.
#Row percent, overall counts, and counts by delivery type are reported.  Column percents can 
#be specified by the row.p=FALSE option.
#By default chi-square tests of independence are performed.

d_type.cattable <- cattable(data=Delivery,
                            vars=c("child.sex", "laceration"),
                            byVar="delivery.type")
 
#For variables listed under "fisher=" Fisher's exact test of independence is performed.
#The reported test statistic is the odds ratio.

d_type.cattable <- cattable(data=Delivery,
                            vars=c("child.sex", "laceration"),
                            fisher=c("child.sex"),
                            byVar="delivery.type")


#All variables listed in a single table

d_type.catconttable <- catconttable(data=Delivery,
                                    vars=c("maternal.age", "ga.weeks", "child.sex", "wt.gram",
                                           "grava", "apgar1", "laceration"),
                                    median=c("grava", "apgar1"),
                                    normal="maternal.age",
                                    fisher="child.sex",
                                    byVar="delivery.type")

#Code for writing ctable objects to a file.  See write.ctable() for more information

#Write to PDF
require(lazyWeave)
lazy.write(
  lazy.file.start(),
  write.ctable(d_type.catconttable),
  lazy.file.end(),
  OutFile="SampleOutput.tex")

#Generate a pdf in the working directory
lazy.build("SampleOutput.tex")

unlink("SampleOutput.tex")
unlink("SampleOutput.pdf")

#Write to HTML
require(lazyHTML)
html_write(
  html_file_start(),
  write.ctable(d_type.catconttable, reportFormat="html"),
  html_file_end(),
  OutFile="SampleOutput.html")

unlink("SampleOutput.html")