# Conf. limits on effect size ### See also CI-RMSSE.R in R-Stuff. # Suppose that F = 6.90 on 4 and 55 df # Giancola study library(MBESS) limits <- conf.limits.ncf(F = 6.90, conf.level = .95, df.1 = 4, df.2 = 55) lower <- limits$Lower.Limit upper <- limits$Upper.Limit cat("The lower and upper confidence limits on the non-centrality parameter (lambda) are \n") cat(lower," and ", upper,"\n") #Not shown in text # We want the root-mean-square standardized effect (RMSSE) #phi <- sqrt(lambda/(k-1)*n) k = 5 n = 12 lambda.1 <- limits$Lower.Limit lambda.2 <- limits$Upper.Limit phi.1 <- sqrt(lambda.1/((k-1)*n)) phi.2 <- sqrt(lambda.2/((k-1)*n)) cat("The lower and upper confidence limits on RMSSE are \n") cat(phi.1," and ",phi.2) #### Rewriting this for the Eysenck one-way example in Exercise 11.1 limits <- conf.limits.ncf(F = 9.08, conf.level = .95, df.1 = 4, df.2 = 45) lower <- limits$Lower.Limit upper <- limits$Upper.Limit lower upper # But this only gives us CI on ncp (lambda) # We want the root-mean-square standardized effect (RMSSE) #phi <- sqrt(lambda/(k-1)*n) k = 5 n = 10 lambda.1 <- limits$Lower.Limit lambda.2 <- limits$Upper.Limit phi.1 <- sqrt(lambda.1/((k-1)*n)) phi.2 <- sqrt(lambda.2/((k-1)*n)) cat("The lower and upper confidence limits on RMSSE are \n") cat(phi.1," and ",phi.2)