csCompare {condir}

Statistically compare CRs towards two CSs

Description

Compare CRs towards two CSs within a frequentist and a Bayesian framework.

Usage

csCompare(
  cs1,
  cs2,
  group = NULL,
  data = NULL,
  alternative = "two.sided",
  conf.level = 0.95,
  mu = 0,
  rscale = 0.707,
  descriptives = TRUE,
  out.thres = 3,
  boxplot = TRUE
)

Arguments

cs1

a numeric vector of values. If the data argument is defined, it can refer to either the column index or the column name of the data object. See Details for more information.

cs2

a numeric vector of values. If the data argument is defined, it can refer to either the column index or the column name of the data object. See Details for more information.

group

column index or name that contain the group data. See Details for more information.

data

numeric matrix or data frame that contains the relevant data.

alternative

a character string for the specification of the alternative hypothesis. Possible values: "two.sided" (default), "greater" or "less".

conf.level

Interval's confidence level.

mu

a numeric value for the mean value or mean difference.

rscale

the scale factor for the prior used in the Bayesian t.test.

descriptives

Returns basic descriptive statistics for cs1 and cs2.

out.thres

The threshold for detecting outliers (default is 3). If set to 0, no outliers analysis will be performed. See Details below for more information.

boxplot

Should a boxplot of the variables be produced (default is TRUE)?

Details

csCompare performs both a student t-test (using the stats::t.test function) and a Bayesian t-test (using the BayesFactor::ttest.tstat). If cs1 and/or cs2 are or refer to multiple columns of a matrix or a data.frame, then the row means are computed before the t-tests are performed. In case group is NULL, paired-samples t-tests will be run. In case the group is different than NULL, then the csCompare first computes difference scores between the cs1 and the cs2 (i.e., cs1 - cs2). In case the group argument is defined but, after removal of NA's (stats::na.omit), only one group is present, a paired samples t-test is run. In case of independent samples t-test, the function runs a Welch's t-test.

Regarding outliers, those are detected based on the deviations from the standardized residuals of each test. For example, in case of a paired-samples t-test, the csCompare function will run an additional regression for detecting deviations (defined in the out.thres argument) from the standardized residuals. The detected outliers are removed from both the frequentists and Bayesian analyses.

Value

The function returns (at least) 3 list objects. These are: descriptives, freq.results, and bayes.results. In case outliers are detected, then the outlier analyses are returned as well with the name res.out as prefix to all list objects. For example, the descriptive statistics of the outlier analyses, can be indexed by using obj$res.out$descriptives, with obj being the object of the csCompare results.

The values of the descriptives are described in psych::describe.

The values of the freq.results are: method: which test was run.

alternative: the alternative hypothesis.

WG1, WG2: the Shapiro test values, separately for group 1 and group 2. In case of a paired-samples t-test, the WG2 is 0.

WpG1, WpG2: the p-values of Shapiro test, separately for group 1 and group 2. In case of a paired-samples t-test, the WpG2 is 0.

null.value: The value defined by mu (see above).

LCI, HCI: The low (LCI) and high (HCI) bounds of the confidence intervals.

t.statistic: Logical.

df: The degrees of freedom of the t-test performed.

p.value: The p-value of the performed t-test.

cohenD: The Cohen's d for the performed t-test.

cohenDM: The magnitude of the resulting Cohen's d.

hedgesG: The Hedge's g for the performed t-test.

hedgesGM: The magnitude of the resulting Hedge's g.

The values of the bayes.results are:

LNI, HNI: The low (LNI) and high (HNI) intervals of the hypothesis to test.

rscale: The used scale (see rscale argument above).

bf10: The BF10.

bf01: The BF01.

propError: The proportional error of the computed Bayes factor.

References

Krypotos, A. M., Klugkist, I., & Engelhard, I. M. (2017). Bayesian hypothesis testing for human threat conditioning research: An introduction and the condir R package. European Journal of Psychotraumatology, 8.

Rouder, J. N., Speckman, P. L., Sun, D., Morey, R. D., & Iverson, G. (2009). Bayesian t-tests for accepting and rejecting the null hypothesis. Psychonomic Bulletin & Review, 16, 225-237

See Also

t.test, ttest.tstat

Examples

set.seed(1000)
csCompare(cs1 = rnorm(n = 100, mean = 10), cs2 = rnorm(n = 100, mean = 9))