Using the classical F test statistic for a balanced one-way design, this function computes the corresponding Bayes factor test.
oneWayAOV.Fstat(F, N, J, rscale = "medium")
F statistic from classical ANOVA
number of observations per cell or group
number of cells or groups
numeric prior scale
For F statistics computed from balanced one-way designs,
this function can be used to compute the Bayes factor
testing the model that all group means are not equal to
the grand mean, versus the null model that all group
means are equal. It can be used when you don't have
access to the full data set for analysis by
lmBF, but you do have the test statistic.
For details about the model, see the help for
anovaBF, and the references therein.
The Bayes factor is computed via Gaussian quadrature.
a vector of length 2 containing the computed log(e) Bayes factor (against the intercept-only null), along with a proportional error estimate on the Bayes factor.
oneWayAOV.Fstat should only be used with F values
obtained from balanced designs.
Richard D. Morey (email@example.com)
Morey, R. D., Rouder, J. N., Pratte, M. S., \& Speckman, P. L. (2011). Using MCMC chain outputs to efficiently estimate Bayes factors. Journal of Mathematical Psychology, 55, 368-378
1 2 3 4 5 6 7 8 9 10 11 12 13
## Example data "InsectSprays" - see ?InsectSprays require(stats); require(graphics) boxplot(count ~ spray, data = InsectSprays, xlab = "Type of spray", ylab = "Insect count", main = "InsectSprays data", varwidth = TRUE, col = "lightgray") ## Classical analysis (with transformation) classical <- aov(sqrt(count) ~ spray, data = InsectSprays) plot(classical) summary(classical) ## Bayes factor (a very large number) Fvalue <- anova(classical)$"F value" result <- oneWayAOV.Fstat(Fvalue, N=12, J=6) exp(result[['bf']])