Suppose I have following results from my 2 experiments that contain 3 Block each and there are 4 treatments per Block (the example is R based)

dft <- data.frame(Exp = rep(1:2, each = 12), 
                 Block = rep(rep(1:3, each = 4), each = 2), 
                 Treat = rep(1:4, 6), 
                 Plot = rep(1:12, 2),
                 y = rnorm(24, mean = 10, sd = 2))

I want to find out if there are differences between Treat and Exp. I used folowing model

myMod <- lme(y ~ Exp*Treat, random = ~1|Block/Plot, data = dft, method = "ML")

but I got the comment that I should include Block nested in Exp in my model I did that too

myMod1 <- lme(y ~ Exp/Block + Exp*Treat, random = ~1|Block/Plot, data = dft, method = "ML")

and than compare both models which were not significantly different

anova(myMod, myMod1)

if there are no differences which model is more logical in this case???

  • $\begingroup$ The simplest one. $\endgroup$ – user2974951 Feb 5 '19 at 8:38

The principle of Occam's Razor, or model parsimony, would apply here. Given two models with similar results, we should prefer the simpler one.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.