any2any: any2any

In ManyLabsOpenScience/manylabRs: Analysis scripts and data from the Many Labs projects

Description

Converts most common test statistics into most common (signed) effect sizes.

Usage

any2any(testInfo, df1 = NULL, df2 = NULL, N = NULL, n1 = NULL,
  n2 = NULL, esType = NA, var.lor = NA, CIcalc = TRUE, CL = 0.95,
  rID = 0, q = 1, alternative = "two", keepDirection = TRUE,
  keepSign = TRUE, keepSignNames = c("r", "l.r", "u.r", "fisher.z",
  "l.z", "u.z"))

Arguments

df1	Degrees of freedom
df2	NULL or degrees of freedom of the denominator for the f-distribution.
N	Number of data points used in calculation of test-statistic.
n1	Number of data points in sample 1.
n2	Number of data points in sample 2.
esType	Type of test statistic. One of: "t", "lm.t", "f", "lm.f", "r", "X2", "Z", "lm.Z"
CIcalc	If TRUE (default) the Confidence Interval for the test statistic in x will be calculated using the "Confidence limits for noncentral parameters" functions in package (e.g., for type - "t": conf.limits.nct).
CL	Confidence Limit (default: .95).
rID	Correlation among predictor values in a linear model.
q	Number of predictors in the model.
alternative	Alternative hypothesis (default = "two").
keepSign	Return effect size with sign of test statistic? (default = TRUE).
keepSignNames	Which effect sizes should keep the sign if keepSign = TRUE? Default is to keep the sign for: "r","l.r","u.r","fisher.z","l.z","u.z".
st	Value(s) of a test statistic.

Details

The procedure to calculate a variety of effect sizes is as follows:

• If CIcalc == FALSE, package::compute.es will be used to convert the test statistic to a large number of effect size estimates. The confidence intervals around the effect size estimates will be based meta-analytic estimates of effect size variance (e.g., for type - "t": tes).

• If CIcalc == TRUE, package::MBESS will be used to calculate the confidence interval for the test statistic based on its noncentral distribution (e.g., for type - "t": conf.limits.nct). Subsequently the test statistic, as well as its lower and upper confidence limit will each be passed to compute.es seperately.

• If keepSign == TRUE the sign of the test statistic will be copied to all the effect sizes in keepSignNames.

Value

The effect sizes calculated by compute.es corresponding to the test statistic(s), with either meta-analytic, or, exact CI.

Note

The prefix "lm" is currently disregarded, but will be implemented in future versions to indicate the test statistic is in fact a fixed factor in a linear model.

Author(s)

Fred Hasselman (inspired by RP:P function any2r by CHJ Hartgerink)

ManyLabsOpenScience/manylabRs documentation built on May 14, 2019, 5:21 p.m.