The Iris Data of Edgar Anderson and Ronald A. Fisher
– Some Background Information on Our Practice Dataset
Edgar Anderson suspected that I. versicolor may be an allopolyploid hybrid: I. versicolor = I. setosa (2n) x I. virginica (4n)
(confirmed by Lim et al. 2007)
This may have supported the establishment of the species, by preventing back-crossing to its parents.
Fisher applied his Linear Discriminant Analysis technique to Anderson’s data, in order to test the hypothesis of additive gene action:
if true, versicolor should be twice as similar to virginica than to sertosa!
# Not Allowed variable names:2myvar <-"John"# starts with a number.2myvar <-"John"# starts with a '.' numbermy-var <-"John"# special charactermy var <-"John"# space_my_var <-"John"# starts with a '_'my_var% <-"John"# special characterTRUE<-"John"# reserved words
Data Types
Data Type
Example
Verify
value
Logical
TRUE / FLASE
x<-TRUEprint(x)class(x)
TRUE logical
Numeric
1.3, 5, 4.2
x<-1.35print(x)class(x)
1.35 numeric
Integer
1L, 0L, 4L
x<-35Lprint(x)class(x)
35 integer
Complex
2+3i
x<-2+3iprint(x)class(x)
2+3i complex
Character
“Hello!”
x<-"Hello!"print(x)class(x)
Hello! character
R Data Structure
The variables are assigned with R-Objects and the data type of the R-object becomes the data type of the variable. There are many types of R-objects. The frequently used ones are −
# Create a vector.apple_colors <-c('green','green','yellow','red','red','red','green')# Create a factor object.factor_apple <-factor(apple_colors)# Print the factor.print(factor_apple)
[1] green green yellow red red red green
Levels: green red yellow
# Create a matrix.M =matrix( c('a','a','b','c','b','a'), nrow =2, ncol =3, byrow =TRUE)# Print the matrixprint(M)
[,1] [,2] [,3]
[1,] "a" "a" "b"
[2,] "c" "b" "a"
# Create an array.a <-array(c('green','yellow'),dim =c(1,3,2))# Print the arrayprint(a)
# Create the data frame.BMI <-data.frame(gender =c("Male", "Male","Female"), height =c(152, 171.5, 165), weight =c(81,93, 78),Age =c(42,38,26))# Print the data frameprint(BMI)
gender height weight Age
1 Male 152.0 81 42
2 Male 171.5 93 38
3 Female 165.0 78 26
Operators
Operators are the symbols that tell the compiler to perform specific mathematical or logical manipulations. R language is rich in built-in operators and provides the following types of operators −
In this session, we’ll try build an understanding of base R functions for data input/output (I/O) and data reshaping using the iris dataset.
Beyond simply running code, we’ll discuss why you might choose one function over another, highlighting their specific strengths and trade-offs, when it makes sense.
We’ll to this in an interactive way.
CSV formats
Base R provides a versatile suite of I/O functions. Some are highly configurable (e.g., read.table()), while others wrap common defaults for convenience (e.g., read.csv()).
When you need full control—different delimiters, quoting rules, or no headers— the generic read.table() and write.table() shine. They accept parameters like sep, quote, na.strings, and more.
When performance, fidelity and reproducibility matter—especially for large objects—R’s binary formats (.RDS / .RData) beat text. saveRDS() / readRDS() handle single objects, while save() / load() manage multiple objects.
tmp <-"iris.RDS"saveRDS(iris, tmp)iris_rds <-readRDS(tmp)save(iris, iris_csv, file ="iris_data.RData")rm(iris, iris_csv)load("iris_data.RData")save.image("workspace.RData") # saves your entire sessionrm(list =ls())load("workspace.RData") # reloads the entire session)
Data Reshaping: Base R Tools
Base R reshaping functions cover pivoting, stacking, grouping, and merging. We’ll compare each pair to understand when to use one versus another.
reshape() handles complex wide⇄long pivots in one call via the direction argument. It’s powerful but can be verbose when specifying varying and times.
stack() / unstack() provide a quick way to collapse or expand multiple columns into key-value pairs, but without the record-level identifiers that reshape() preserves.
Result: Many-to-many join: e.g. 25 versicolor in each → 625 rows.
Why?merge() pairs every matching row for duplicated keys.
# Create per-species indexdir1$idx <-1:nrow(dir1)dir2$idx <- (50+1):(50+nrow(dir2))# Now merge on both Species & idxmatched <-merge(dir1, dir2, by =c("Species","idx"), all=T)matched# Drop idx if you likematched$idx <-NULL
Outcome: Exactly one-to-one pairing, recovering the intended alignment.
Try It: Shuffle rows of dir2 before merging—do you still get correct matches? Why? Challenge: Perform a full outer join (all=TRUE) on (Species, idx) and inspect NAs.
4. Summary & Takeaways
cbind() / rbind(): Great for straightforward stacking when dimensions align exactly; no key matching.
merge(): Key-based alignment; duplicates produce Cartesian products unless you add an ID for one-to-one matching.
By consciously adding IDs when joining on duplicated keys, you ensure your merged table mirrors your intended relational structure—no surprises!
There are easier ways to run a function over the columns of a table – tomorrow!
R Plots (base) – some preparations
Finally we define our own coloring scheme:
## These colors are supposed to be easy to## discriminate for sight-impaired people.species_colors =setNames( ## makes a named vectorpalette()[1:3], rownames(species_means) )species_colors
Barplots represent sign and absolute value of numbers by the direction and length of bars.
If called with a matrix as first argument, the function produces one plot for each column:
barplot(## one plot per column == trait,## one bar == species mean! species_means,## do not stack the barsbeside=TRUE,## larger group labelscex.names=1.5,col=species_colors, )## add a legend (plot "augmentation"!)legend("topright",rownames(species_means), fill=species_colors,cex=1.5)
R Plots (base) – barplot()
If we want to plot trait means per species, we must change the rows of matrix species_means (= the species) into columns, because barplot() reads a matrix by column.
This is done by the t() function ("transpose"):
m <-t(species_means) ## TRANSPOSEbarplot(## one plot per column == species,## one bar == trait mean! m,## do not stack the barsbeside=TRUE,## larger group labelscex.names=2,col=trait_colors,## increase y limit to fit the legendylim =c(0,10),cex =2 )## add a legend (plot "augmentation"!)legend(x=1,y=10, ##"topright",rownames(m), fill=trait_colors)
R Plots (base) – pie()
Piecharts are a quick-and-dirty alternative for representing numbers.
The pie() function can only represent one set of numbers at a time. In addition, comparing angles on a piechart is visually not as easy as comparing bar heights.
pie(species_means[,"Petal.Length"], labels =rownames(species_means),main =paste("Mean Petal Lengths in ","Fisher's Iris Species"),col=species_colors,cex=2, ## larger text annotationcex.main =2## larger title)
R Plots (base) – pie()
Piecharts are a quick-and-dirty alternative for representing numbers.
The pie() function can only represent one set of numbers at a time. In addition, comparing angles on a piechart is visually not as easy as comparing bar heights.
pie(species_means["setosa",], labels =colnames(species_means),main =paste("Mean Flower Organ Dimensions","in Iris setosa"),col=trait_colors,cex=2, ## larger text annotationcex.main =2## larger title)
R Plots (base) – plot()
The plot() function is an extremely versatile workhorse for x/y plots.
As an “initializing” function, it may be called to just create an empty canvas, to be filled later:
## (see par() for graphical parameters!)plot(x=NULL,y=NULL,## Note that if you start empty, ## you have to set the canvas ## dimensions yourself!xlim=c(0,3),ylim=c(0,9),xlab ="x dimension",ylab ="y dimension")
R Plots (base) – plot()
Or it is called with an initial set of data, with the option to extend the plot later:
plot(x=iris$Petal.Width, y=iris$Petal.Length)
R Plots (base) – plot()
There is more than one way to relate x/y dimensions to data!
## Initialize a new plot, ## using formula notation to specify x and y:plot(Petal.Length ~ Petal.Width,data = iris)
R Plots (base) – plot()
Add a grid:
grid()
R Plots (base) – plot()
Overplot some points with color, in order to identify a group in your data:
points(Petal.Length ~ Petal.Width, data=subset(iris, Species =="versicolor" ),pch=21, # symbol code 21: bullet with# separate interior color (bg) # and border color (col)## See points()!## Set color manually:col="red", bg ="red")
R Plots (base) – plot()
Color all points by species, using our named vector species_colors:
plot(Petal.Length ~ Petal.Width, data=iris,pch=21,## index the "species_colors" vector## by species names:col=species_colors[Species], bg =species_colors[Species] )legend("topleft",rownames(species_means),fill=species_colors)
R Plots (base) – plot()
Points with adjacent positions in the input can be connected by lines, using different line styles. A typical use case is a line graph, with x as a running number or ID.
## See par() for line-related parameters!## Make a new data.frame, ## containing only setosa:df <-subset(iris, Species=="setosa")plot(# x is now the row number in dfx=1:nrow(df),xlab="individual plant",y=df$Petal.Width,ylab="Petal.Width",## show both points and ## connecting lines:type="b",## line width:lwd =2,## line style = dashed:lty=2,main="Iris setosa")
R Plots (base) – plot()
It can make sense to connect some points in a general scatterplot by lines.
The augmenting function lines() can do this.
Here, we want to connect the (x,y) means of our three species:
## Initialize a plain x/y plot:plot(Petal.Length ~ Petal.Width,data=iris)## Add colored mean points, connected by lineslines(x=species_means[,"Petal.Width"],y=species_means[,"Petal.Length"],type="b", # show both points and lines pch=21,bg=species_colors,cex=2.5)
R Plots (base) – plot()
Annotate individual points:
text(x=species_means[,"Petal.Width"], y=species_means[,"Petal.Length"],labels=rownames(species_means),pos=4, ## put labels to the right of points## (see ?text)cex=3## expansion factor for the text)
R Plots (base) – plot()
Function abline() adds indicator lines to a plot.
Regression line:
## Initialize a plain x/y plot:plot(Petal.Length ~ Petal.Width,data=iris,## Put a title:main ="Regression Line Example")## Mark the linear regression line:abline(lm(Petal.Length ~ Petal.Width,data=iris ),lty=2, lwd=2,col="blue" )
R Plots (base) – plot()
Function abline() adds indicator lines to a plot.
Lines marking locations or slopes of interest:
## Initialize a plain x/y plot:plot(Petal.Length ~ Petal.Width,data=iris,main ="abline() example" )## Horizontal and vertical markers:abline(h =2.5, col ="red", lwd=2## line width )abline(v =0.75, col ="yellow", lwd=2)## An "assumed" regression line for reference:abline(a=1,b=2,lwd=2)
R Plots (base) – layout()
Several plots can be combined on the same page in a grid-like layout.
The grid is specified by a matrix of possible plot positions, like so:
## prepare the layout matrixm <-matrix(1:4, nrow=2, ncol=2, byrow=FALSE)m
[,1] [,2]
[1,] 1 3
[2,] 2 4
The first plot will go to grid position 1, the second to position 2 … .
The hist() function is one of those “dual use functions”:
With add=FALSE, it initializes the device and the coordinate system, while
with add=TRUE, its output goes directly to an existing plot.
setosa <-subset(iris,Species=="setosa")versicolor <-subset(iris,Species=="versicolor")virginica <-subset(iris,Species=="virginica")## Plot the histogram of setosa, ## and initialize the entire plot: hist(setosa$Petal.Length,col=species_colors["setosa"],add=FALSE, ## this is the default## initialize to full x range !xlim=range(iris$Petal.Length),## full y range you usually ## only know after some trials ..ylim=c(0,22),## x-axis labelxlab="Petal Length",## larger axis labels:cex.lab =2,main ="Petal Length Distributions",## larger title:cex.main =2)
R Plots (base) – hist()
The hist() function is one of those “dual use functions”:
With add=FALSE, it initializes the device and the coordinate system, while
with add=TRUE, its output goes directly to an existing plot.
The hist() function is one of those “dual use functions”:
With add=FALSE, it initializes the device and the coordinate system, while
with add=TRUE, its output goes directly to an existing plot.
hist(virginica$Petal.Length,add=TRUE,col=species_colors["virginica"])legend(x=2,y=22, legend=c("setosa","versicolor","virginica"), fill=species_colors,cex =1.5, ## larger script in legend )
R Plots (base) – boxplot()
The boxplot() function has “dual-use” capabilities, too.
However it can accept a formula with a factor on the right hand side, and it will split the dataset automatically according to the factor levels. So we can plot all species at once:
boxplot(Petal.Length ~ Species, data = iris,## colors are not automatically## inferred from the factor levels!col = species_colors )
R Plots (base) – boxplot()
Let’s add a boxplot for the global Petal.Length distribution (all species merged):
# Repeat the last boxplot, with an extended x axis: boxplot(Petal.Length ~ Species, data = iris, col = species_colors,xlim=c(0,6) )
boxplot(iris$Petal.Length, # take the entire column!add=TRUE, at=5, ## position on x axisnames="all species",show.names=TRUE )
R Plots (base) – Saving Plots From RStudio
R Plots (base) – Saving Plots From RStudio
Digression: Software is Usually Built From Bits and Pieces
… they come under the names of subroutines, macros, functions …
… in code, they are used like ’commands’:
Actually the function name invokes a piece of hidden code
… hiding complexity
… yet allowing to easily access complex algorithms
