dgrtwo / dgrtwo.github.com Goto Github PK

I'm having an issue running the example from "Understanding empirical Bayesian hierarchical modeling (using baseball statistics)"
http://varianceexplained.org/r/hierarchical_bayes_baseball/

I run everything before and up to:

crossing(bats = c("L", "R"),
         AB = c(10, 100, 1000, 10000)) %>%
  augment(fit2, newdata = .) %>%
  mutate(H = .3 * AB,
         alpha0 = .fitted / sigma,
         beta0 = (1 - .fitted) / sigma,
         alpha1 = alpha0 + H,
         beta1 = beta0 + AB - H,
         estimate = alpha1 / (alpha1 + beta1),
         conf.low = qbeta(.025, alpha1, beta1),
         conf.high = qbeta(.975, alpha1, beta1),
         record = paste(H, AB, sep = " / ")) %>%
  ggplot(aes(estimate, record, color = bats)) +
  geom_point() +
  geom_errorbarh(aes(xmin = conf.low, xmax = conf.high)) +
  labs(x = "Estimate w/ 95% credible interval",
       y = "Batting record",
       color = "Batting hand")

and get:

Error in mutate_impl(.data, dots) : 
  Evaluation error: non-numeric argument to binary operator.

I figured out the error is coming from the fact that sigma is a function, and so is causing the calculation of alpha0 to fail. I'm not too familiar with tidyr, and so am not sure how to pull out sigma for the predicted/fitted new data...

I appreciate your help in advance!

Hello David,
I'm using the minimal mistakes theme for my blog
Could you advice on how did you create the "Subscribe" feature in your blog.
Thank you for your time.

Cheers
Ashish

I could not figure out how to use inflate function properly...

When I reproduce what you implemented, I've got the following errors.

http://varianceexplained.org/r/bayesian_ab_baseball/

two_players %>%
+ broom::inflate(x = seq(.28, .33, .00025)) %>%
+ mutate(density = dbeta(x, alpha1, beta1)) %>%
+ ggplot(aes(x, density, color = name)) +
+ geom_line() +
+ labs(x = "Batting average", color = "")
Error in mutate_impl(.data, dots) :
Evaluation error: object 'alpha1' not found.

Hey David, fan of your site. I'm trying to use Minimal Mistakes as well - was wondering how you left-aligned the masthead links?

Thanks.

I enjoyed your post on "Examining the arc of 100,000 stories" - thanks! For me the associated RMarkdown file raises the following error:

Line 204: Error in FUN(X[[i]], ...) : object 'decile' not found

It looks like this is a result of joining two tables containing the same decile column name.