This function performs a contrast test on a given dataset to compare the means of multiple variables using weighted contrasts. It calculates several statistics, including the t-value, F-value, p-value, confidence intervals, and effect sizes (Cohen's f and partial eta squared).
Usage
contrast_test(
dat,
weight = NULL,
paired = FALSE,
alternative = c("two.sided", "less", "greater"),
conf.level = 0.95,
verbose = TRUE
)Arguments
- dat
A data frame or tibble containing the data to be analyzed. Each column represents a variable. For unbalanced data, please fill missing values with NA.
- weight
A numeric vector of weights to apply to each variable in the contrast. The sum of the weights must be zero.
- paired
A logical indicating whether the data is paired (TRUE) or unpaired (FALSE). Default is FALSE.
- alternative
A character string specifying the type of test. Possible values are "two.sided" (default), "less", or "greater".
- conf.level
A numeric value indicating the confidence level for the confidence intervals. Default is 0.95.
- verbose
Logical. If
TRUE, prints additional messages.
Value
A data frame containing the following columns:
- weights
The weights applied to each variable in the contrast.
- estimate
The weighted contrast estimate.
- lower
The lower bound of the confidence interval.
- upper
The upper bound of the confidence interval.
- t
The t-value for the contrast.
- F
The F-value for the contrast.
- df_error
The degrees of freedom for the error.
- SS_contrast
The sum of squares for the contrast component.
- SS_error
The sum of squares for the error.
- SS_effect
The sum of squares for the effect.
- SS_total
The toal sum of squares for the data.
- p
The p-value for the contrast test.
- eta2
The eta squared effect size.
- peta2
The partial eta squared effect size.
- cohens_f
The Cohen's f effect size.
- n_total
The total number of observations in the unpaired data.
- n_total_na
The number of missing values in the unpaired data.
- n_pair
The number of non-missing pairs in the paired data.
- n_pair_na
The number of missing pairs in the paired data.
Details
This function performs a contrast analysis on a given dataset using the provided weights. It computes statistics such as the t-value, F-value, confidence intervals, and effect sizes (Cohen's f and partial eta squared).
If the data is paired, pair-wise exclusion is applied. The function supports two-tailed and one-tailed tests depending on the value of the alternative parameter.
See also
trend_test() for trend analysis using contrast tests.
Examples
# Example of linear trend analysis with unpaired data
dat <- data.frame(A = rnorm(100), B = rnorm(100), C = rnorm(100))
weight <- c(1, -1, 0)
contrast_test(dat, weight)
#>
#> -----------------------------------------------
#> Contrast analysis for independent samples
#> -----------------------------------------------
#> Mean = ( 0.07079, 0.11486, -0.03005)
#> n = (100, 100, 100)
#> Weight = (1, -1, 0)
#> -----------------------------------------------
#> Contrast = -0.04407
#> CI = [-0.3332, 0.2451]
#> -----------------------------------------------
#> t = -0.2999, df = 297, p = 0.7644363
#> (F = 0.08996)
#> -----------------------------------------------
#> SS_contrast = 0.0971, SS_error = 320.56
#> SS_total = 321.66, SS_effect = 1.104
#> -----------------------------------------------
#> eta squared = 0.0003019
#> partial eta squared = 0.0003028, Cohen's f = 0.0174
#> -----------------------------------------------
#> Note:
#> The CI represents the frequentist confidence interval.
#> (Confidence level = 0.95)
# Example with paired data
dat <- data.frame(A = rnorm(50), B = rnorm(50), C = rnorm(50))
weight <- c(1, -2, 1)
contrast_test(dat, weight, paired = TRUE)
#>
#> -----------------------------------------------
#> Contrast analysis for paired samples
#> -----------------------------------------------
#> Mean = ( 0.1699, 0.2073, -0.2158)
#> N = 50
#> Weight = (1, -2, 1)
#> -----------------------------------------------
#> Contrast = -0.4605
#> CI = [-1.136, 0.2145]
#> -----------------------------------------------
#> t = -1.354, df = 98, p = 0.1788597
#> (F = 1.833)
#> -----------------------------------------------
#> SS_contrast = 1.767, SS_error = 94.48
#> SS_total = 158.83, SS_effect = 64.35
#> -----------------------------------------------
#> eta squared = 0.01113
#> partial eta squared = 0.01836, Cohen's f = 0.1368
#> -----------------------------------------------
#> Note:
#> The CI represents the frequentist confidence interval.
#> (Confidence level = 0.95)