conclusion.Rmd

# (PART\*) Part V: Conclusion {-}

# Summary

Implementing Bayesian methods has never been easier

<span class="pack">rstanarm</span> and <span class="pack">brms</span> allow one to worry about the concepts instead of syntax

Bayesian analysis provides additional ways to:

- combat complex models
- diagnosis issues
- get creative with results

Even standard/traditional/basic models become more fun!




# Exercise

Whether a survey respondent agrees or disagrees with a conservative statement about the role of women in society

- 1974/1975 Survey
- Women should take care of running their homes and leave running the country up to men
- Modeled as a function of the gender and years of education (0-20) of the respondents
- Data from the <span class="pack">HSAUR3</span> package
- Observations are aggregate counts at education levels for men and women



```{r basic_logistic_data_prep, echo=FALSE, eval=FALSE}
data("womensrole", package = "HSAUR3")
womensrole = womensrole %>% 
  mutate(total = agree + disagree,
         agree50 = agree/total >= .5)
readr::write_csv(womensrole, 'data/womensrole.csv')
```

Run a binomial logistic regression modeling the proportion of those who agreed 
- If you are more familiar with binary logistic regression, you may 'unrole' this data to be disagree-agree for each individual (the analysis is the same)


```{r basic_logistic}
data("womensrole", package = "HSAUR3")

# alternative not requiring package
womensrole = readr::read_csv('https://github.com/m-clark/easy-bayes/raw/master/data/womensrole.csv')

womensrole_glm <- glm(cbind(agree, disagree) ~ education + gender,
                        data = womensrole, 
                        family = binomial)
summary(womensrole_glm)
```

Now use <span class="pack">brms</span> or <span class="pack">rstanarm</span> to run the same analysis

Explore it fully by:

- plotting the coefficients (<span class="func">stanplot</span>)
- use <span class="pack">shinystan</span> to explore diagnostics (<span class="func">launch_shinystan</span>)
- explore the posterior predictive distribtution (<span class="func">pp_check</span>)
- plot the marginal effect of the predictor variables (<span class="func">marginal_effects</span>)

```{r basic_logistic_result}
my_model_bayes = ?
```