Essential Statistics with R: Cheat Sheet
Important libraries to load
If you don’t have a particular package installed already: install.packages(Tmisc).
library(readr) # for optimized read with read_csv() instead of read.csv()
library(dplyr) # for filter(), mutate(), %>%, etc. see dplyr lesson.
library(ggplot2) # for making plots in this lesson
library(broom) # OPTIONAL: for model tidying with tidy(), augment(), glance()
library(Tmisc) # OPTIONAL: for gg_na() and propmiss()
The pipe: %>%
When you load the
dplyr
library you can use
%>%
, the pipe. Running
x %>% f(args)
is the same as
f(x, args)
. If you wanted to
run function
f()
on data
x
, then run function
g()
on that, then run function
h()
on that result: instead of nesting multiple functions,
h(g(f(x)))
, it’s preferable and more readable to create a chain or pipeline of functions:
x %>% f %>% g %>% h
. Pipelines can be
spread across multiple lines, with each line ending in
%>%
until the pipeline terminates. The keyboard shortcut for inserting
%>%
is
Cmd+Shift+M on Mac, Ctrl+Shift+M on Windows.
Functions
Function Description
read_csv("path/nhanes.csv") Read nhanes.csv in the path/ folder (readr)
View(df) View tabular data frame df in a graphical viewer
head(df) ;tail(df) Print first and last few rows of data frame df
mean,median,range Descriptive stats. Remember na.rm=TRUE if desired
is.na(x) Returns TRUE/FALSE if NA.sum(is.na(x)) to count NAs
filter(df, ..,) Filters data frame according to condition ... (dplyr)
t.test(y~grp, data=df) T-test mean yacross grp in data df
wilcox.test(y~grp, data=df) Wilcoxon rank sum / Mann-Whitney Utest
lmfit <- lm(y~x1+x2, data=df) Fit linear model yagainst two x’s
anova(lmfit) Print ANOVA table on object returned from lm()
summary(lmfit) Get summary information about a model fit with lm()
TukeyHSD(aov(lmfit)) ANOVA Post-hoc pairwise contrasts
xt <- xtabs(~x1+x2, data=df) Cross-tabulate a contingency table
addmargins(xt) Adds summary margin to a contingency table xt
prop.table(xt) Turns count table to proportions (remember margin=1)
chisq.test(xt) Chi-square test on a contingency table xt
fisher.test(xt) Fisher’s exact test on a contingency table xt
mosaicplot(xt) Mosaic plot for a contingency table xt
factor(x, levels=c("wt", "mutant")) Create factor specifying level order
relevel(x, ref="wildtype") Re-level a factor variable
glm(y~x1+x2, data=df, family="binomial") Fit a logistic regression model
power.t.test(n, power, sd, delta) T-test power calculations
power.prop.test(n, power, p1, p2) Proportions test power calculations
tidy() augment() glance() Model tidying functions in the broom package
ggplot2 basics
Build a plot layer-by-later, starting with a call to
ggplot()
, specifying the data and aesthetic mappings, for instance, to x/y
coordinates and color. Continue building a plot by adding layers such as geometric objects (geoms) or statistics, like a trendline.
The example below will use
mydata
, plot
xvar
and
yvar
on the xand yaxes, plot points colored by levels of
groupvar
, and add a
linear model trendline.
ggplot(mydata, aes(xvar, yvar)) +geom_point(aes(color=groupvar)) +geom_smooth(method="lm")