What is MBESS?

Methods for the Behavioral, Educational, and Social Sciences (MBESS) implements methods that are especially useful to researchers working within the behavioral, educational, and social sciences (both substantive researchers and methodologists). Many of the methods contained within MBESS are applicable to quantitative research in general.

The Goal of MBESS

The goal of MBESS is to provide substantive and quantitative researchers, especially in the behavioral, educational, and social sciences, with a package that contains useful functions for the unique types of quantitative techniques used in these domains. Although there are many R packages developed with specific areas of inquiry in mind, few packages exist that are devoted to the idiosyncratic techniques used within the behavioral, educational, and social sciences sciences (Of course, many of these techniques also extend to other domains.). One of the long term goals of MBESS is to contain a relatively complete set of functions to compute confidence intervals for various effect sizes (especially those based on noncentral distributions). Another long term goal is to contain a relatively complete set of functions to compute necessary sample size from the Power Analytic and the Accuracy in Parameter Estimation approach.

Contribute

Anyone is free to contribute to this project. Contributions can be as simple as submitting ideas for functions that you would like to see included in the MBESS package (feel free to email Ken Kelley, the package author, with any ideas) or by submitting functions (hopefully with documentation) that can be included in the package (of course credit will be given for individuals who contribute code).

Selected Functions Contained within MBESS

Many of the functions contained within MBESSare related to my methodological work in one way or another. Below is a listing of some of the function names (in italics) and a very brief description of what they do. After MBESS in installed and loaded into the R session (using library(MBESS) from within R), a question mark followed by the function name (?FunctionName) will return the help page for particular function. As can be seen, many of the functions relate to sample size planning (from the accuracy in parameter estimation and the power analytic approach) and confidence intervals for effect sizes.

The function ci.smd() can be used to calculate the confidence limits for the population standardized mean difference (i.e., what is generally referred to as Cohen’s d, with Delta being that population quantity) using the square root of the pooled variance as the divisor.

For example, two groups of 10 participants each have a standardized mean difference of 1.25 with the corresponding t-value of 2.7951. To form a 95% confidence interval for the population value, the function can be specified as follows.

> ci.smd(ncp=2.7951, n.1=10, n.2=10, conf.level=.95)
$Lower.Conf.Limit.smd
[1] 0.2700283

$smd
[1] 1.250007

$Upper.Conf.Limit.smd
[1] 2.201482

Suppose the population standardized mean difference is believed to be .5. To obtain a 95% confidence interval whose width is expected to be no wider than .30, the necessary sample size per group can be planned as follows.

> ss.aipe.smd(delta=.5, conf.level=.95, width=.30)
[1] 353

To ensure with .90 probability that the 95% confidence interval width obtained in a particular study is no wider than .30:

> ss.aipe.smd(delta=.5, conf.level=.95, width=.30, assurance=.90)
[1] 357

# Exchangable correlation structure
Rho.YX <- c(.3, .3, .3, .3, .3)
RHO.XX <- rbind(
c(1, .5, .5, .5, .5), 
c(.5, 1, .5, .5, .5), 
c(.5, .5, 1, .5, .5),
c(.5, .5, .5, 1, .5),
c(.5, .5, .5, .5, 1))

ss.aipe.src(width=.1, which.width="Full", sigma.Y=1, sigma.X=1, RHO.XX=RHO.XX,
Rho.YX=Rho.YX, which.predictor=1)
[1] "Necessary sample size such that the expected 95 confidence interval using noncentral methods is 0.1 is 2191"
[1] 2191