How many random factor levels and replicates do you suggest to carry out a single random factor design when you want a 95% confidence interval for the common error variance, and you want the interval width to be about 1/3 of the common error variance? What if there were two random factors say, A and B? How many replicates, and levels of each random factor do you recommend to obtain a 95% confidence interval with width 1/3 of the common error variance