Description
Please submit both pdf and Rmd files (or html and Rmd files).
Part II: Split/Apply/Combine and tidyverse warm-up
This famous (Fisher’s or Anderson’s) iris data set gives the measurements in centimeters of the variables
sepal length and width and petal length and width, respectively, for 50 flowers from each of 3 species of iris.
The species are Iris setosa, versicolor, and virginica.
Consider the following loop that computes the mean of each quantitative variable split by species and stores
the computed means in a matrix named MeanFlowers.
# define a matrix of zeros
MeanFlowers <- matrix(0,nrow=4,ncol=3)
# define a character vector corresponding to the numeric variable names
measurements <- c(“Sepal.Length”,”Sepal.Width”,”Petal.Length”,”Petal.Width”)
# name the rows and columns of the matrix MeanFlowers
rownames(MeanFlowers) <- measurements
colnames(MeanFlowers) <- c(“setosa”,”versicolor”,”virginica”)
# Loop
for (j in measurements) {
#– R code goes here —-
MeanFlowers[j,] <- round(tapply(iris[,j],iris[,”Species”],mean),4)
}
MeanFlowers
## setosa versicolor virginica
## Sepal.Length 5.006 5.936 6.588
## Sepal.Width 3.428 2.770 2.974
## Petal.Length 1.462 4.260 5.552
## Petal.Width 0.246 1.326 2.026
Problem 1
Replicate the above loop using the Split/Apply/Combine model with base R commands.
Solution goes below
1
## solution goes here —
Problem 2
Repeat question 1 by constructing a pipe, including the split() function from base R and map_df() from
the purrr package.
Solution goes below
## solution goes here —
Part II: More tidyverse with CDC cancer data
Consider the Center of Disease Control data set BYSITE_new.csv, which describes the incidence and
mortality counts of several types of cancer over time. The variables of interest are: YEAR, RACE, SITE,
EVENT_TYPE, COUNT and POPULATION.
Problem 3
Load in the dataset BYSITE_new.csv using the appropriate function from the readr package. Display
the dimension of the cancer tibble.
Solution goes below
## solution goes here —
Base R code for reference.
# Base R code for reference
cancer <- read.csv(“BYSITE_new.csv”,header=T)
dim(cancer)
## [1] 44982 7
Problem 4
Using Base R or tydyverse functions, identify any strange symbols that are recorded in the COUNT variable.
Once you have identified the symbols, use functions from the dplyr package to remove any rows in the cancer
tibble containing these symbols and then convert COUNT to a numeric mode.
## solution goes here —
Problem 5
For a specific tumor and population, a crude rate is calculated by dividing the number of new cancers observed
during a given time period by the corresponding number of people in the population at risk. For cancer, the
result is usually expressed as an annual rate per 100,000 persons at risk. https://ci5.iarc.fr/ci5plus/pages/
glossary.aspx
In reference to our data, this quantity can be calculated by:
CRUDE RATE = 100000 ∗
COUNT
POPULATION
2
Using relevant functions from the dplyr package, create a new variable in your dataframe (or tibble) called
CRUDE_RATE. Then using base R graphics or ggplot, create a histogram of CRUDE_RATE. Note
that the crude rates are not bounded between [0,1] because they are calculated per 100,000 persons at risk.
## solution goes here —
Problem 6
Compute the average incidence rate of prostate cancer for each level of RACE. To solve this problem,
students must build a pipe (magrittr package) and utilize the appropriate functions from the dplyr package.
Also compare your results to a base R solution. Include both the tidyverse and base R solutions in your final
write-up. Note: before computing the average incidence rates, students should filter the data as follows:
i. Extract the rows corresponding to EVENT_TYPE level Incidence
ii. Extract the rows corresponding to SITE level Prostate
iii. Extract the rows corresponding to SEX level Male
iv. Remove the rows corresponding to YEAR level 2010-2014
v. Remove the rows corresponding to RACE level All Races
Solution goes below
First filter the dataset:
## solution goes here —
Compute the average incidence rate of prostate cancer for each level of RACE.
## solution goes here —
Problem 7
Create a plot in base R or ggplot that shows the incidence rate (CRUDE_RATE) as a function of time
(YEAR), split by the levels of RACE. Make sure to include a legend and label the graphic appropriately.
Before constructing the graphic, perform the data wrangling tasks using a pipe and functions from the dplyr
package, i.e., the same filtering tasks from problem 6. Students can use some base R functions in the pipe if
needed and the plotting code can be included inside or outside the pipe.
Solution goes below
## solution goes here —
Problem 8
Fit five simple linear regression models, one for each level of RACE, relating the incidence rate
(CRUDE_RATE) as a function of time (YEAR). Collect the estimated slopes, t-statistics and p-values of
your estimated models. The collection of slopes describe whether cancer has increased or decreased over the
selected time period and the p-values describe if the increase or decrease is statistically significant. Solve this
problem using a pipe and functions from the dplyr and purrr packages. Note: use the same filtered data
from problem 4 and problem 4 in this analysis.
Some hints: (i) this exercise is a natural extension of problem 7; (ii) if needed, students can also define
their own functions used in the pipe; (iii) students are not required to use a single pipe to solve this question
but it’s a fun challenge if interested.
3
Solution goes below
## solution goes here —
4
