ind.oneway.second: A one-way design with independent samples using published...
In rpsychi: Statistics for psychiatric research

Description Usage Arguments Details Value Author(s) References See Also Examples

ind.oneway.second conducts a one-way design with independent samples, namely one-way randomized-group analysis of variance, using published work.

1 2	ind.oneway.second(m, sd, n, unbiased = TRUE, contr = NULL, sig.level = 0.05, digits = 3)

`m`	a numeric vector contains the means (length(`m`) >= 2)
`sd`	a numeric vector contains the sample/unbiased standard deviations (length(`sd`) >= 2)
`n`	a numeric contains the sample size (length(`n`) >= 2)
`unbiased`	`sd` contains unbiased standard deviations (`unbiased` = TRUE, default) or sample standard deviations (`unbiased` = FALSE)
`contr`	a matrix or vector contains the contrast weights
`sig.level`	a numeric contains the significance level (default 0.05)
`digits`	the specified number of decimal places (default 3)

This function conducts a one-way design with independent samples, namely one-way randomized-group analysis of variance, using published work. If you do not specify contr, all possible pairwise contrasts will be calculated. Statistical power is calculated using the following specifications:

(a) small (η^2 = 0.01), medium (η^2 = 0.06), and large (η^2 = 0.14) population effect sizes, according to the interpretive guideline for effect sizes by Cohen (1992)

(b) sample size specified by n

The returned object of ind.oneway.second contains the following components:

`anova.table`	returns a ANOVA table containing sums of squares, degrees of freedom, mean squares, F values
`omnibus.es`	returns a omnibus effect size which is a η^2, and its' confidence interval
`raw.contrasts`	returns raw mean differences, their confidence intervals, and standard errors
`standardized.contrasts`	returns standardized mean differences for the contrasts (Hedges's g), their approximate confidence intervals for population standardized mean differences, and standard errors
`power`	returns statistical power for detecting small (η^2 = 0.01), medium (η^2 = 0.06), and large (η^2 = 0.14) population effect sizes

Yasuyuki Okumura
Department of Social Psychiatry,
National Institute of Mental Health,
National Center of Neurology and Psychiatry
yokumura@blue.zero.jp

Cohen B (2000) Calculating a factorial ANOVA from means and standard deviations. Understanding Statistics, 1, 191-203.

Cohen J (1992) A power primer. Psychological Bulletin, 112, 155-159.

Kline RB (2004) Beyond significance testing: Reforming data analysis methods in behavioral research. Washington: American Psychological Association.

ind.oneway, samplesize.etasq

##Kline (2004) Table 6.3
dat <- data.frame(y = c(9,12,13,15,16,
                       8,12,11,10,14,
                       10,11,13,11,15),
                  x =  rep(factor(c("a","b","c")), each=5)
                  )                 

##contrast 1: a - c, contrast 2: 1/2(a + c) - b
my.cont <- matrix(c(1,0,-1,1/2,-1,1/2), ncol=3, nrow=2, byrow=TRUE)


ind.oneway.second(m = tapply(dat$y, dat$x, mean),
                  sd = tapply(dat$y, dat$x, sd),
                  n= tapply(dat$y, dat$x, length)) 

ind.oneway.second(m = tapply(dat$y, dat$x, mean),
                  sd = tapply(dat$y, dat$x, sd),
                  n= tapply(dat$y, dat$x, length),
                  contr = my.cont)