Lowered the iteration count to balance core count per Claude advice

"Run more shorter chains instead of fewer long ones. For example, 8 chains × 1000 iterations will have less variance in finish times than 4 chains × 2000 iterations, and you get the same total draws. The load-balancing improves because the law of large numbers works in your favor."

Author: migurski

2025/planscore_models_congress.R

@@ -82,8 +82,8 @@ m <- brm(bf(dem_share_imputed ~ dpres_mn + incumb +
                            group="stateabrev"),
                  set_prior("student_t(3, 0.02, 0.05)", class="sd",coef="incumb",
                            group="stateabrev")),
-         cores=detectCores(), chains=4, control=list(adapt_delta=0.99999, max_treedepth=12),
-         warmup=2000, iter=6000, refresh=10, thin=16)
+         cores=detectCores(), chains=detectCores(), control=list(adapt_delta=0.99999, max_treedepth=12),
+         warmup=ceiling(4*2000/detectCores()), iter=ceiling(4*6000/detectCores()), refresh=10, thin=16)
 proc.time() - start
 saveRDS(m, "full_model_2025B_incumbency_congress.rds")
 
@@ -135,7 +135,7 @@ m <- brm(bf(dem_share_imputed ~ dpres_mn +
                  set_prior("student_t(3, 0.07, 0.1)", class="sd",coef="dpres_mn",
                            group="stateabrev")),
          cores=detectCores(), chains=detectCores(), control=list(adapt_delta=0.99999, max_treedepth=12),
-         warmup=2000, iter=6000, refresh=10, thin=16)
+         warmup=ceiling(4*2000/detectCores()), iter=ceiling(4*6000/detectCores()), refresh=10, thin=16)
 proc.time() - start
 saveRDS(m, "full_model_2025B_openseat_congress.rds")
 
@@ -233,7 +233,7 @@ m <- brm(bf(dem_share_imputed ~ dpres_mn + incumb +
                  set_prior("student_t(3, 0.02, 0.05)", class="sd",coef="incumb",
                            group="stateabrev")),
          cores=detectCores(), chains=detectCores(), control=list(adapt_delta=0.99999, max_treedepth=12),
-         warmup=2000, iter=6000, refresh=10, thin=16)
+         warmup=ceiling(4*2000/detectCores()), iter=ceiling(4*6000/detectCores()), refresh=10, thin=16)
 proc.time() - start
 saveRDS(m, "full_model_2025A_incumbency_congress.rds")
 
@@ -285,7 +285,7 @@ m <- brm(bf(dem_share_imputed ~ dpres_mn +
                  set_prior("student_t(3, 0.07, 0.1)", class="sd",coef="dpres_mn",
                            group="stateabrev")),
          cores=detectCores(), chains=4, control=list(adapt_delta=0.99999, max_treedepth=12),
-         warmup=2000, iter=6000, refresh=10, thin=16)
+         warmup=ceiling(4*2000/detectCores()), iter=ceiling(4*6000/detectCores()), refresh=10, thin=16)
 proc.time() - start
 saveRDS(m, "full_model_2025A_openseat_congress.rds")
 

2025/planscore_models_state_leg.R

@@ -74,7 +74,7 @@ m <- brm(bf(dem_share_imputed ~ dpres_mn + incumb +
                  set_prior("student_t(3, 0.02, 0.05)", class="sd",coef="incumb",
                            group="stateabrev")),
          cores=detectCores(), chains=detectCores(), control=list(adapt_delta=0.99999, max_treedepth=12),
-         warmup=2000, iter=6000, refresh=10, thin=16) #warmup and iter for number of cycles
+         warmup=ceiling(4*2000/detectCores()), iter=ceiling(4*6000/detectCores()), refresh=10, thin=16) #warmup and iter for number of cycles
 proc.time() - start
 saveRDS(m, "full_model_2026A_incumbency_statelege.rds")
 
@@ -129,7 +129,7 @@ m <- brm(bf(dem_share_imputed ~ dpres_mn +
                  set_prior("student_t(3, 0.11, 0.1)", class="sd",coef="dpres_mn",
                            group="stateabrev")),
          cores=detectCores(), chains=detectCores(), control=list(adapt_delta=0.99999, max_treedepth=12),
-         warmup=2000, iter=6000, refresh=10, thin=16) #warmup and iter for number of cycles
+         warmup=ceiling(4*2000/detectCores()), iter=ceiling(4*6000/detectCores()), refresh=10, thin=16) #warmup and iter for number of cycles
 proc.time() - start
 saveRDS(m, "full_model_2026A_openseat_statelege.rds")