How do fixed effects in brms::fixedef(brms.fit) differ from fixed effects reported in summary(brms.fit)?

Interfaces brms
1

EDIT: No issue here! Just a stupid mistake.

I have a brms.fit object “puco.full”. When I run:

brms::fixef(puco.full, probs=c(0.05, 0.95))

it returns:

> fixef(puco.full, probs=c(0.05, 0.95))
                Estimate Est.Error        Q5        Q95
Intercept     19960.4330  4761.425 12142.365 27797.8991
HOMO.p         8821.6201  3935.104  2346.349 15294.0025
Trail           763.6816  5299.004 -7941.894  9450.3825
road_dens.sc  -2499.8360  1754.979 -5371.706   376.2785
trail_dens.sc  1088.5278  1849.576 -1948.042  4125.0580
lunar.frac.sc   365.2649  1713.789 -2451.644  3176.5037
CR_CLOSURE.sc -1687.5067  1722.609 -4519.080  1136.3774

but when i run

summary(puco.full)

it returns:

> summary(puco.full)
 Family: gaussian 
  Links: mu = identity; sigma = identity 
Formula: noct ~ HOMO.p + Trail + road_dens.sc + trail_dens.sc + lunar.frac.sc + CR_CLOSURE.sc 
   Data: PUCO (Number of observations: 46) 
Samples: 4 chains, each with iter = 1e+05; warmup = 5000; thin = 1;
         total post-warmup samples = 380000

Population-Level Effects: 
              Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
Intercept     19960.43   4761.43 10568.23 29325.91 1.00   409647   296411
HOMO.p         8821.62   3935.10  1050.12 16588.01 1.00   341705   290486
Trail           763.68   5299.00 -9671.60 11221.62 1.00   365641   291449
road_dens.sc  -2499.84   1754.98 -5962.39   964.08 1.00   369722   289606
trail_dens.sc  1088.53   1849.58 -2554.26  4732.38 1.00   339163   288769
lunar.frac.sc   365.26   1713.79 -3007.93  3742.60 1.00   404041   291533
CR_CLOSURE.sc -1687.51   1722.61 -5083.54  1707.09 1.00   403517   286610

Family Specific Parameters: 
      Estimate Est.Error l-95% CI u-95% CI Rhat Bulk_ESS Tail_ESS
sigma 11086.53   1280.42  8923.49 13927.97 1.00   332236   270779

Samples were drawn using sampling(NUTS). For each parameter, Bulk_ESS
and Tail_ESS are effective sample size measures, and Rhat is the potential
scale reduction factor on split chains (at convergence, Rhat = 1).

It seems the Q5 and Q95 columns of the fixef() call differ from the l-95% CI and u-95% CI columns of summary(). Can anyone explain what the differences between these are? I was under the impression that both would be 95% credible intervals, so they should be identical, but maybe I’m missing something?

2

Maybe because … your fixef() uses a 90% credible interval whereas the summary() uses a 95% credible interval!

Try:

brms::fixef(puco.full, probs=c(0.025, 0.975))

3

Yikes. Sorry for that. Been a long day. I thought I was going crazy.