I am see the relationship between the betas (all specified in the design matrix), but I can’t see where the r_* parameters in the BRMS model are coming from and those are the essential ones.
BRMS_Mod[["exclude"]]
[1] "ordered_Intercept" "fixed_Intercept" "theta"
[4] "chol_cor" "bQ" "hs_global"
[7] "hs_local" "hs_slab" "zb"
[10] "hs_localsp" "zbsp" "Intercept"
[13] "first_Intercept" "merged_Intercept" "zcar"
[16] "nszcar" "zerr" "z_1"
[19] "L_1" "Cor_1" "r_1"
[22] "z_2" "L_2" "Cor_2"
[25] "r_2"
attr(,"save_ranef")
[1] TRUE
attr(,"save_mevars")
[1] FALSE
attr(,"save_all_pars")
[1] FALSE
exclude suggests all that r_1, r_2, z_1, and z_2 are excluded, but tidybayes returns the important variables prefixed by r_
get_variables(BRMS_Mod)
[1] "b_Intercept"
[2] "b_ConditionIntervention"
[3] "b_Session_TypeInitial_Homecare"
[4] "b_Session_TypeInitial_Debrief"
[5] "b_Session_TypeFollowMup_Homecare"
[6] "b_Session_TypeFollowMup_Debrief"
[7] "b_FacilitatorFacilitator2"
[8] "b_ConditionIntervention:Session_TypeInitial_Homecare"
[9] "b_ConditionIntervention:Session_TypeInitial_Debrief"
[10] "b_ConditionIntervention:Session_TypeFollowMup_Homecare"
[11] "b_ConditionIntervention:Session_TypeFollowMup_Debrief"
[12] "b_Session_TypeInitial_Homecare:FacilitatorFacilitator2"
[13] "b_Session_TypeInitial_Debrief:FacilitatorFacilitator2"
[14] "b_Session_TypeFollowMup_Homecare:FacilitatorFacilitator2"
[15] "b_Session_TypeFollowMup_Debrief:FacilitatorFacilitator2"
[16] "sd_Condition__Intercept"
[17] "sd_Condition:Group__Intercept"
[18] "r_Condition[Control,Intercept]"
[19] "r_Condition[Intervention,Intercept]"
[20] "r_Condition:Group[Control_10,Intercept]"
[21] "r_Condition:Group[Control_12,Intercept]"
[22] "r_Condition:Group[Control_20,Intercept]"
[23] "r_Condition:Group[Control_36,Intercept]"
[24] "r_Condition:Group[Control_4,Intercept]"
[25] "r_Condition:Group[Control_40,Intercept]"
[26] "r_Condition:Group[Control_6,Intercept]"
[27] "r_Condition:Group[Control_8,Intercept]"
[28] "r_Condition:Group[Intervention_1,Intercept]"
[29] "r_Condition:Group[Intervention_13,Intercept]"
[30] "r_Condition:Group[Intervention_17,Intercept]"
[31] "r_Condition:Group[Intervention_19,Intercept]"
[32] "r_Condition:Group[Intervention_29,Intercept]"
[33] "r_Condition:Group[Intervention_3,Intercept]"
[34] "r_Condition:Group[Intervention_33,Intercept]"
[35] "r_Condition:Group[Intervention_41,Intercept]"
[36] "prior_Intercept"
[37] "prior_b"
[38] "prior_sd_Condition__Intercept"
[39] "prior_sd_Condition:Group__Intercept"
[40] "lp__"
[41] "accept_stat__"
[42] "stepsize__"
[43] "treedepth__"
[44] "n_leapfrog__"
[45] "divergent__"
[46] "energy__"
I see the betas in the stan model variables, but the indexing looks quite different and that’s what I’m struggling with.
get_variables(Stan_Fit)
[1] "b[1]" "b[2]" "b[3]"
[4] "b[4]" "b[5]" "b[6]"
[7] "b[7]" "b[8]" "b[9]"
[10] "b[10]" "b[11]" "b[12]"
[13] "b[13]" "b[14]" "Intercept"
[16] "sd_1[1]" "z_1[1,1]" "z_1[1,2]"
[19] "sd_2[1]" "z_2[1,1]" "z_2[1,2]"
[22] "z_2[1,3]" "z_2[1,4]" "z_2[1,5]"
[25] "z_2[1,6]" "z_2[1,7]" "z_2[1,8]"
[28] "z_2[1,9]" "z_2[1,10]" "z_2[1,11]"
[31] "z_2[1,12]" "z_2[1,13]" "z_2[1,14]"
[34] "z_2[1,15]" "z_2[1,16]" "r_1_1[1]"
[37] "r_1_1[2]" "r_2_1[1]" "r_2_1[2]"
[40] "r_2_1[3]" "r_2_1[4]" "r_2_1[5]"
[43] "r_2_1[6]" "r_2_1[7]" "r_2_1[8]"
[46] "r_2_1[9]" "r_2_1[10]" "r_2_1[11]"
[49] "r_2_1[12]" "r_2_1[13]" "r_2_1[14]"
[52] "r_2_1[15]" "r_2_1[16]" "b_Intercept"
[55] "prior_b" "prior_Intercept" "prior_sd_1_1"
[58] "prior_sd_2_1" "lp__" "accept_stat__"
[61] "stepsize__" "treedepth__" "n_leapfrog__"
[64] "divergent__" "energy__"
Is something behind the scenes in BRMS combining r_1_1 and r_2_1 parameters to create r_* variables indexed from r_variable[r_1_1, r_2_1] ?
I realize this is like a basic/pedestrian series of questions, but I’m not sure where to look to get an understanding of the stan code.