Advanced Statistics and Data Science (ABCD) CHEAT SHEET

Word Equations Summary Tables Simple Statistics

outcome = explanatory + other stuff # compute five-number summary mean(data_set$Y)
favstats(~ Y, data = data_set) var(data_set$Y)
Y = X + other stuff sd(data_set$Y)
# create frequency table
Basics tally(data_set$Y) cohensD(Y ~ X, data = data_set)
tally(~ Y, data = data_set) cor(Y ~ X, data = data_set)
print("Hello world!")
# tally by condition b1(Y ~ X, data = data_set)
# assign value to object tally(~ Y < 1900, data = data_set) b1(one_model)
my_number <- 5
# two-way frequency table pre(Y ~ X, data = data_set)
# combine values into vector tally(Y ~ X, data = data_set, margin = TRUE, f(Y ~ X, data = data_set)
my_vector <- c(1, 2, 3) format = “proportion”)
# sample F for X2
# first element in vector f(Y ~ X1 + X2,
my_vector[1] data = data_set, predictor = ~X2)

# orders values or cases

sort(my_vector) Data Frame
# arithmetic operations # structure of data frame # arrange rows by variable
sum(1, 2, 100), +, -, *, / str(data_set) arrange(data_set, Y)
sqrt(157)
abs(data_set$Y) # view first/last six rows # creates data frame from csv file
head(data_set) data_set <- read.csv("file_name", header = TRUE)
# logical operations tail(data_set)
>, <, >=, <=, ==, !=, |, & # convert quantitative variable
# select multiple variables # to categorical
# results in a new variable with values select(data_set, Y1, Y2) factor(data_set$Y)
# of TRUE or FALSE factor(data_set$Y,
data_set$C <- data_set$A > data_set$B # first six rows of selected variables levels = c(1,2),
head(select(data_set, Y1, Y2)) labels = c("A", "B"))

Probability Distribution # transform values

# select variable (a column) recode(data_set$Y, "0" = 0, "1" = 50, "2" = 100)
# calculate the probability area
xpnorm(65.1, data_set$mean, data_set$sd) data_set$Y
# creates two equal sized groups
# find rows that meet condition ntile(data_set$Y, 2)
zscore(data_set$Y)
data_set[data_set$Y > 40] # convert categorical variable
# returns t at this probability filter(data_set, Y > 300) # to quantitative
qt(.975, df = 999) as.numeric(data_set$Y)
# returns F at this probability # find rows that do not have NA
qf(.95, df1 = 1, df2 = 100) filter(data_set, is.na(Y) == FALSE)
filter(data_set, !is.na(Y))
# CI using t distribution
confint(empty_model)

# calculate p-value using F-distribution

xpf(sample_f, df1 = 2 , df2 = 10)

Page: 1 ▷ Updated: 2024-10 ▷ Learn more about CourseKata @ https://coursekata.org

Advanced Statistics and Data Science (ABCD) CHEAT SHEET
Simulation Fitting and Evaluating Models
# sample without replacement # randomize sampling distribution of b1s, # empty model
sample(data_set, 6) # centered on 0 empty_model <- lm(Y ~ NULL,
sdob1 <- do(1000) * data = data_set)
# sample with replacement b1(shuffle(Y) ~ X, data = data_set)
resample(data_set, 10) # use one explanatory variable
# bootstrap sampling distribution of b1s, one_model <- lm(Y ~ X, data = data_set)
do(3) * resample (data_set, 10) # centered on sample b1
sdob1_boot <- do(1000) * # use more than one explanatory variable
# mixes up values in a variable b1(Y ~ X, data = resample(data_set)) # multivariate model
shuffle(data_set$Y) multi_model <- lm(Y ~ X1 + X2, data = data_set)
# count the number of b1s at the upper
# simulate sampling 10000 Ys # and lower extreme # all the model comparisons that can be
# from normal distribution tally(sdob1$b1 > sample_b1 | # made in relation to the multivariate model
sim_Y <- rnorm(10000, Y_stats$mean, sdob1$b1 < -sample_b1) generate_models(multi_model)
Y_stats$sd)
# model predictions and residuals
# put simulated Ys into dataframe # return TRUE for middle 95% of distribution data_set$empty_predict <- predict(empty_model)
data_set<- data.frame(sim_Y) middle(sdob1$b1, .95) data_set$empty_resid <- resid(empty_model)
# randomize sampling distribution of PREs # produce ANOVA table
# simulate
sdopre <- do(1000) * pre(shuffle(Y) ~ X, anova(empty_model)
# sampling distribution of means data = data_set)
sdom_sim <- do(10000) * mean(rnorm(157, supernova(one_model)
Y_stats$mean, Y_stats$sd)) # randomize sampling distribution of Fs
sdof <- do(1000) * # t-test, using pooled variance
# bootstrap f(shuffle(Y) ~ X, data = data_set) t.test(Tip ~ Condition, data = data_set,
# sampling distribution of means var.equal=TRUE)
sdom_boot <- do(10000) * # counts extreme Fs
mean(resample(data_set$Y, 157)) sample_f <- f(shuffle(Y) ~ X, data = data_set) # pairwise comparison corrections:
tally(~f > sample_f, data = sdof) # "Tukey","Bonferroni","none"
pairwise(one_model, correction = "none")

Page: 2 ▷ Updated: 2024-10 ▷ Learn more about CourseKata @ https://coursekata.org

Advanced Statistics and Data Science (ABCD) CHEAT SHEET
Visualizations
gf_histogram(~ Y, data = data_set) %>% # sampling distribution of b1
gf_point(Y ~ X, data = data_set) # change labels gf_histogram(~b1, data = sdob1,
gf_labs(title = "Graph Title", fill = ~middle(b1, .95)) %>%
x = "Y_Name", y = "Frequency") # modify the limits on x- and y-axes
gf_lims(x = c(-12, 12), y = c(0, 70))

gf_jitter(Y ~ X, data = data_set)

# faceted grid of histograms
gf_histogram(~ Y, data = data_set) %>%
gf_facet_grid(X ~ .)

gf_point(Y ~ X, data = data_set) %>%

# add model predictions as red points
gf_point(Y ~ X , shape = 1, size = 3,
color = "firebrick") %>%
# add best fitting model as a red line
gf_boxplot(Y ~ X, data = data_set) gf_model(one_model, color = “red”)

gf_dhistogram(~ Y, data = data_set,

fill = "orange") %>%
gf_density()

gf_boxplot(Y ~ X, data = data_set, fill = pairwise(one_model, plot = TRUE)

"orange") %>%
gf_jitter(height = 0, alpha = .2, size = 3)
gf_bar( ~ Y, data = data_set)

Page: 3 ▷ Updated: 2024-10 ▷ Learn more about CourseKata @ https://coursekata.org