Microbial Informatics

• Project 1 is due today
• Read The Art of R Programming (Chapters 6 and 7)

• How to design a small function from pseudocode
• Additional details about vectors and matrices

• Lists
• Data frames
• Tables
• Factors

• a data structure that can combine different data types (eg. numeric, vectors, strings, etc)
• useful for returning complex data structures from functions
• important for organizing data: imagine creating a list for each gene in a genome - gene name, coordinates, length, function, DNA sequence, translated sequence, a reference to experiments, etc...
• Could have a vector of lists... or a data frame...

ps <- list(name=c("Pat", "Schloss"), position="Associate Professor", time.at.um=5)
ps$name

## [1] "Pat"     "Schloss"

ps$position

## [1] "Associate Professor"

ps$time.at.um

## [1] 5

str(ps)

## List of 3
##  $ name      : chr [1:2] "Pat" "Schloss"
##  $ position  : chr "Associate Professor"
##  $ time.at.um: num 5

length(ps)

## [1] 3

names(ps)

## [1] "name"       "position"   "time.at.um"

unlist(ps)

##                 name1                 name2              position 
##                 "Pat"             "Schloss" "Associate Professor" 
##            time.at.um 
##                   "5"

is.vector(unlist(ps))

## [1] TRUE

ps$position

## [1] "Associate Professor"

ps[["position"]]

## [1] "Associate Professor"

ps[[2]]

## [1] "Associate Professor"

ps[2]

## $position
## [1] "Associate Professor"

ps$kids.names <- c("mary", "patrick", "joe", "john", "ruth", "jacob", "peter")
ps$kids[3]

## [1] "joe"

ps[[4]][3]

## [1] "joe"

ps$married <- TRUE

ps

## $name
## [1] "Pat"     "Schloss"
## 
## $position
## [1] "Associate Professor"
## 
## $time.at.um
## [1] 5
## 
## $kids.names
## [1] "mary"    "patrick" "joe"     "john"    "ruth"    "jacob"   "peter"  
## 
## $married
## [1] TRUE

ps$position <- NULL
ps

## $name
## [1] "Pat"     "Schloss"
## 
## $time.at.um
## [1] 5
## 
## $kids.names
## [1] "mary"    "patrick" "joe"     "john"    "ruth"    "jacob"   "peter"  
## 
## $married
## [1] TRUE

data <- list(x=1:10, y=100:110)
lapply(data, median)    #returns a list

## $x
## [1] 5.5
## 
## $y
## [1] 105

sapply(data, median)    #returns a vector

##     x     y 
##   5.5 105.0

x <- runif(100)
y <- c(rep("red", 50), rep("blue", 50))

t <- t.test(x~y)
str(t)

## List of 9
##  $ statistic  : Named num 0.105
##   ..- attr(*, "names")= chr "t"
##  $ parameter  : Named num 97.6
##   ..- attr(*, "names")= chr "df"
##  $ p.value    : num 0.916
##  $ conf.int   : atomic [1:2] -0.107 0.119
##   ..- attr(*, "conf.level")= num 0.95
##  $ estimate   : Named num [1:2] 0.471 0.465
##   ..- attr(*, "names")= chr [1:2] "mean in group blue" "mean in group red"
##  $ null.value : Named num 0
##   ..- attr(*, "names")= chr "difference in means"
##  $ alternative: chr "two.sided"
##  $ method     : chr "Welch Two Sample t-test"
##  $ data.name  : chr "x by y"
##  - attr(*, "class")= chr "htest"

t$statistic

##         t 
## 0.4470755

t$parameter

##       df 
## 97.01836

t$p.value

## [1] 0.655817

t$estimate

## mean in group blue  mean in group red 
##          0.4656561          0.4411594

• like a matrix, but each column can have a different mode
• We saw these in the first several classes when you were looking at the metadata file
• A data frame is really a multidimensional list

name <- c("Schloss", "Young" , "Mobley", "SwansonJ")
rank <- c("Asst", "Assoc", "Chair", "Full")
cool <- c(10, 1, 9, 8)
directory <- data.frame(name, rank, cool)
directory

##       name  rank cool
## 1  Schloss  Asst   10
## 2    Young Assoc    1
## 3   Mobley Chair    9
## 4 SwansonJ  Full    8

directory[,1]

## [1] Schloss  Young    Mobley   SwansonJ
## Levels: Mobley Schloss SwansonJ Young

directory[,"name"]

## [1] Schloss  Young    Mobley   SwansonJ
## Levels: Mobley Schloss SwansonJ Young

directory[1,]

##      name rank cool
## 1 Schloss Asst   10

directory[directory$name=="Schloss",]

##      name rank cool
## 1 Schloss Asst   10

rownames(directory) <- directory$name
directory["Schloss",]

##            name rank cool
## Schloss Schloss Asst   10

• Vector that can contain only predefined values and is used to store categorical data.
  

x <- factor(c("a", "b", "b", "c", "d"))
x

## [1] a b b c d
## Levels: a b c d

str(x)

##  Factor w/ 4 levels "a","b","c","d": 1 2 2 3 4

levels(x)

## [1] "a" "b" "c" "d"

x[2] <- "e"

## Warning in `[<-.factor`(`*tmp*`, 2, value = "e"): invalid factor level, NA
## generated

x

## [1] a    <NA> b    c    d   
## Levels: a b c d

x <- factor(c("a", "b", "b", "c", "d"), levels=c("a", "b", "c", "d", "e"))
x

## [1] a b b c d
## Levels: a b c d e

levels(x)

## [1] "a" "b" "c" "d" "e"

table(x)

## x
## a b c d e 
## 1 2 1 1 0

x <- factor(c("a", "b", "b", "c", "d"), levels=c("a", "e", "b", "c", "d"))
x

## [1] a b b c d
## Levels: a e b c d

levels(x)

## [1] "a" "e" "b" "c" "d"

table(x)

## x
## a e b c d 
## 1 0 2 1 1

• splits a vector into groups

set.seed(2)
ages <- sample(20:40, 20, replace=TRUE)
gender <- factor(sample(c("Female", "Male"), 20, replace=TRUE))
pol <- factor(sample(c("D", "R", "I"), 20, replace=TRUE))
split(ages, gender)

## $Female
##  [1] 34 23 39 39 22 37 31 31 25 24 21
## 
## $Male
## [1] 23 32 29 35 23 28 37 40 29

• generate factors based on pre-defined bins

range <- seq(20,40,5)
segments <- cut(ages, range)    #(20,30]: 20 < x <= 30
table(segments)

## segments
## (20,25] (25,30] (30,35] (35,40] 
##       5       7       5       2

• or let cut figure out the breaks for you

segments <- cut(ages, 5)
table(segments)

## segments
##   (20,23.6] (23.6,27.2] (27.2,30.8] (30.8,34.4]   (34.4,38] 
##           1          10           2           4           3

• or make equally abundant groups

range <- quantile(ages)
segments <- cut(ages, range)
table(segments)

## segments
##   (20,25] (25,26.5] (26.5,33]   (33,38] 
##         5         4         7         3