Hello,

I have been trying to use brms for parameter estimation with the following model:

```
anim.diff$full <- brm(formula = y ~ param1*param2*param3*param4 +
(1 | sub),
data = dat,
family = student(),
warmup = 1000, iter = 5000,
cores = parallel::detectCores(),
chains = 4, control = list(adapt_delta = .99),
prior = prior,
save_pars = save_pars(all = TRUE))
```

Where I get a posterior for some of the parameters with a high mass above/below zero, with e.g., P(param < 0) = 0.96. However, when I then use loo to compare to a model without the respective parameter, the model without it is always the winning model. In fact, an intercept-only model seems to perform better than a model with any combination of the four predictors, while bayes_R2 of the full model including all predictos is quite a lot larger than the bayes_R2 of the intercept-only model (R2 = 0.25 vs R2 = 0.04).

Iām now trying to find out whether this is because my model is somehow misspecified, or whether it really means my model does in fact have very low out of sample predictive power.

Loo shows me the effective number of parameters (p_loo) is very high (suggesting bad misspecification), while pareto-k values are ok:

```
Computed from 16000 by 107 log-likelihood matrix
Estimate SE
elpd_loo 21.7 7.8
p_loo 19.8 2.2
looic -43.5 15.6
------
Monte Carlo SE of elpd_loo is 0.1.
Pareto k diagnostic values:
Count Pct. Min. n_eff
(-Inf, 0.5] (good) 103 96.3% 3020
(0.5, 0.7] (ok) 4 3.7% 3048
(0.7, 1] (bad) 0 0.0% <NA>
(1, Inf) (very bad) 0 0.0% <NA>
All Pareto k estimates are ok (k < 0.7).
See help('pareto-k-diagnostic') for details.
```

But I canāt seem to figure out why/how my model is misspecified.

This is the fit of my posterior predictive distribution:

as well as mean recovery:

plotting loo_pit shows some diversion from a uniform distribution:

N.B. my outcome variable ranges from -1 to 1, where only certain values are possible (calulated based on change in number of correct responses out of 8 trials).

I would be very grateful if anyone could offer some insight on how I should interpret these results & diagnostics. Thanks!