WalkerHarrison

library(tidyverse)
library(ggvis)

## this column needs to be unique for each row (primary key)!
mtcars$url <- paste0("https://en.wikipedia.org/wiki/", 1:nrow(mtcars))

mtcars %>%
  ggvis(~mpg, ~hp) %>%
  layer_points(size := 50, size.hover := 200,
               fillOpacity := 0.4, fillOpacity.hover := 0.7,

WalkerHarrison

# pitch/play data scraped with pitchRx

pitches18 %>% 
  filter(type == 'X', 
         pitch_type == "FF") %>%
  mutate(TB = case_when(
    event == "Single" ~    1,
    event == "Double" ~    2,
    event == "Triple" ~    3,
    event == "Home Run" ~  4,

WalkerHarrison

git clone https://github.com/jeetsukumaran/Syrupy
cd Syrupy
python setup.py install
syrupy.py -m 20

WalkerHarrison

strapz <- list()
k <- 10
for (i in 1:k){
  meps.bs <- meps %>% sample_n(nrow(.), replace = TRUE)
  prop.score.fit <- multinom(race ~ . - healthExp, data = meps.bs, maxit = 200,
                             trace = FALSE, model = TRUE)
  strapz[[i]] <- prop.score.fit
}

save(strapz, file = "strapz.Rdata")

WalkerHarrison

library(expm)

###### polynomial ######

p <- 4
mt <- rep(1, p)
G <- diag(p); G[1:(p-1), 2:p] <- G[1:(p-1), 2:p] + diag(p-1);
F <- c(1, rep(0, p-1))

print(G); print(mt); print(F)

WalkerHarrison

library(expm)
p <- 4
mt <- rep(1, p)
G <- diag(p); G[1:(p-1), 2:p] <- G[1:(p-1), 2:p] + diag(p-1);
F <- c(1, rep(0, p-1))

print(G); print(mt); print(F)

k <- 5
G %^% k

WalkerHarrison

library(tidyverse)

phis <- c(0.6, -0.4, 0.2, 0.2, 0.1)
p <- length(phis)
X <- rnorm(p)

G <- rbind(phis, 
           cbind(diag(p-1),
                 rep(0, p-1)))

WalkerHarrison

    figure(4); clf
    subplot(2,1,1)
    h=sqrt(squeeze(sC(itrend,itrend,:)))'.*sq; ciplot(sm(itrend,:)-h,sm(itrend,:)+h,1:T,[.85 .85 .85]); hold on
    plot(1:T,sm(itrend,:),'b-'); eval(xa); hold off
    title('90% smoothed posterior intervals for trend trend '); ylabel('Parameter')

WalkerHarrison

def build_hash_table(C, d, k, data):
    vals = np.arange(C*d)
    HT = defaultdict(set)
    HTR = defaultdict(set)
    I = np.random.choice(vals, k, replace = False) # figure out if replace = True or False

    for j in range(n):
            # for every point, generate hashed point and sample k bits
        p = data[j]
        hashed_point = lsh_hash(p, C)[I]

WalkerHarrison

a = 3
b = 4
sigma = 2

1 - pnorm((a-b)/sigma)
pnorm((b-a)/sigma)