I have a list of patient samples that were tested at 3 clinical sites; for site 1, reagent lots 1 and 2 were applied; for site 2, reagent lots 1 and 3 were applied; for site 3, reagent lots 2 and 3 were applied. So this is a partially crossed design. Which mixed model I should use or actually both can be used:

Model 1: Y= Mean+ SITE + LOT + LOT*SITE + Error, OR Model 2: Y= Mean+ SITE + LOT (SITE) + Error,

I will appreciate any feedback!!



If you want to know how to run the model in R with package lme4 you would use specify the crossed random effects as (1|Lot) + (1|Site) where the slopes are fixed and intercepts are random.

I believe the equation is


where Yij = the outcome

γ00 = the grand mean (intercept of fixed effects)

u0j = intercept for the first random effect (R1)

u1j = intercept for the second random effect (R2)

y10 = coefficient for the first fixed effect X1

y20 = coefficient for the second fixed effect X2

y30 = coefficient for the third fixed effect X3

This equation is of course if you have predictors as fixed effects and random crossed effects.

| cite | improve this answer | |
  • $\begingroup$ Yij=(γ00+u0j+u1j)+γ10X1ij+γ20X2ij+γ30X3ij+eij is equal to Yij=(γ00+uj)+γ10X1ij+γ20X2ij+γ30X3ij+eij ? because I cannot find a way to separate u0j and u1j. $\endgroup$ – user158565 Oct 14 '18 at 19:55
  • $\begingroup$ If you specify your random effects as crossed, say in lme4 (1|var1) + (1|var2) then in the output you will get the u0j and u1j instead of a single uj. $\endgroup$ – Kreitz Gigs Oct 14 '18 at 21:43
  • $\begingroup$ u0j and u1j are the intercepts of the random effects. $\endgroup$ – Kreitz Gigs Oct 15 '18 at 0:58
  • $\begingroup$ then should use u_j and u_k. $\endgroup$ – user158565 Oct 15 '18 at 3:56

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.