This section presents a data set that shows how different data types should be read into R as well as several functions that are useful for working with different types of R objects. Consider the data stored as a CSV file at
https://raw.githubusercontent.com/alanarnholt/Data/master/POPLAR3.CSV
The following description of the data is from Minitab 15:
The data from Poplar3.CSV is read into the data frame poplar using the read.csv() function, and the first five rows of the data frame are shown using the function head() with the argument n = 5 to show the first five rows of the data frame instead of the default n = 6 rows.
When dealing with imported data sets, it is always good to examine their contents using functions such as str() and summary(), which show the structure and provide appropriate summaries, respectively, for different types of objects.
From typing str(poplar) at the R prompt, one can see that all seven variables are either integer or numeric. From the description, the variables Site and Treatment are factors. Further investigation into the experiment reveals that year and Age are factors as well. Recall that factors are an extension of vectors designed for storing categorical information. The results of summary(poplar) indicate the minimum values for Diameter, Height, and Weight are all -99, which does not make sense unless one is told that a value of -99 for these variables represents a missing value. Once one understands that the variables Site, Year, Treatment, and Age are factors and that the value -99 has been used to represent missing values for the variables Diameter, Height, and Weight, appropriate arguments to read.csv() can be entered. The data is now read into the object poplarC using na.strings = "-99" to store the NA values correctly. The argument colClasses= requires a vector that indicates the desired class of each column.
In the event different values (999, 99, 9999) for different variables (var1, var2, var3) are used to represent missing values in a data set, the argument na.strings= will no longer be able to solve the problem directly. Although one can pass a vector of the form na.strings = c(999, 99, 9999), this will simply replace all values that are 999, 99, or 9999 with NAs. If the first variable has a legitimate value of 99, then it too would be replaced with an NA value. One solution for this problem in general is to read the data set into a data frame (DF), to assign the data frame to a different name so that the cleaned up data set is not confused with the original data, and to use filtering to assign NAs to values of var1, var2, and var3 that have entries of 999, 99, and 999, respectively.
Once a variable has its class changed from int to factor, labeling the levels of the factor can be accomplished without difficulties. To facilitate analysis of the poplarC data, labels for the levels of the variables Site and Treatment are assigned.
readr functionsdata.table function fread()Basic scatterplot:
ggplot(data = poplarC,
mapping = aes(x = diameter, y = height)) +
geom_point() 
#
ggplot(data = poplarC,
mapping = aes(x = diameter, y = height, color = treatment)) +
geom_point() +
theme_bw() +
geom_smooth(se = FALSE)
#
ggplot(data = poplarC,
mapping = aes(x = diameter, y = height, color = treatment)) +
geom_point() +
theme_bw() +
geom_smooth(se = FALSE) +
facet_grid(cols = vars(age), rows = vars(site))