No worries! I should have been more clear. :)

Did you include the prior specification in your first run?

This is all very well documented in the `rstanarm`

vignettes. If you remove `prior_count=dirichlet(1)`

, it would not really change anything, since the default prior *is* `dirichlet(1)`

. Also, you can can obviously change the R^2 prior location: The more explanatory power you think your predictors have (jointly), the higher should be your prior R^2. You can also specify `prior=NULL`

which will result in flat priors (uniform on the real number line) for the regression coefficients. This is generally not advisable.

You can verify that the default prior for the counts is indeed \text{Dirichlet(1)} and that `prior=NULL`

yields flat priors:

```
fit_default <- stan_polr(
FWS1 ~ Ethnic1 + Fam1 + Eco1 + Health1 + Safety1 + Community1 + Religios1 + Housing1,
data = df,
method = "logistic",
prior = NULL,
init_r = 0.1,
seed = 12345,
algorithm = "sampling"
)
```

Now call `prior_summary`

on the `fit_default`

object:

```
> prior_summary(fit_default)
Priors for model 'fit_default'
------
Coefficients
~ flat
Counts
~ dirichlet(concentration = [1,1,1])
------
See help('prior_summary.stanreg') for more details
```

You should really check out the `rstanarm`

vignettes. They are awesome!