Task 3: Learning by Doing: Explore a New Dataset on Your Own

Task 10 is meant to be a litte more advanced and combines many aspects you have learnt today.

1. download the dataset mlb.RData from ILIAS and load it into your project

load("../data/mlb.RData")
ls()

## [1] "mlb"

1. How many records are their in the dataset, what are the columns of dataset

nrow(mlb)

## [1] 1034

1. Find the tallest player

head(mlb)

##            player team       position height weight   age
## 1   Adam_Donachie  BAL        Catcher     74    180 22.99
## 2       Paul_Bako  BAL        Catcher     74    215 34.69
## 3 Ramon_Hernandez  BAL        Catcher     72    210 30.78
## 4    Kevin_Millar  BAL  First_Baseman     72    210 35.43
## 5     Chris_Gomez  BAL  First_Baseman     73    188 35.71
## 6   Brian_Roberts  BAL Second_Baseman     69    176 29.39

mlb[which.max(mlb$height), ]

##        player team       position height weight   age
## 929 Jon_Rauch  WAS Relief_Pitcher     83    260 28.42

# in cm
cm(mlb[which.max(mlb$height), "height"])

## [1] 210.8

1. Create a copy of the data called mlb_eu_measure and transform lbs to kg and inches to cm. Hint: use cm() and the fact that lbs*0.45359237 is one kg.

mlb_eu_measure <- mlb
lbs_to_kg <- function(lbs, b = 0.45359237) lbs * b
mlb_eu_measure$weight <- lbs_to_kg(mlb_eu_measure$weight)
mlb_eu_measure$height <- cm(mlb_eu_measure$height)

1. Find the heaviest player for each position!

# okay let's split the dataset first using position as a factor
pos <- split(mlb_eu_measure, mlb_eu_measure$position)
lapply(pos, function(x) x[which.min(x$weight), ])

## $Catcher
##          player team position height weight  age
## 796 Carlos_Ruiz  PHI  Catcher  182.9  77.11 28.1
## 
## $Designated_Hitter
##         player team          position height weight   age
## 53 Jerry_Owens  CWS Designated_Hitter  190.5  88.45 26.03
## 
## $First_Baseman
##            player team      position height weight   age
## 76 Howie_Kendrick  ANA First_Baseman  177.8  81.65 23.64
## 
## $Outfielder
##           player team   position height weight   age
## 14 Brandon_Fahey  BAL Outfielder    188  72.57 26.11
## 
## $Relief_Pitcher
##           player team       position height weight   age
## 828 Fabio_Castro  PHI Relief_Pitcher  172.7  68.04 22.11
## 
## $Second_Baseman
##            player team       position height weight   age
## 457 Alexi_Casilla  MIN Second_Baseman  175.3  72.57 22.61
## 
## $Shortstop
##              player team  position height weight   age
## 288 Oswaldo_Navarro  SEA Shortstop  182.9  68.04 22.41
## 
## $Starting_Pitcher
##           player team         position height weight   age
## 368 Odalis_Perez   KC Starting_Pitcher  182.9  68.04 29.73
## 
## $Third_Baseman
##            player team      position height weight   age
## 80 Maicer_Izturis  ANA Third_Baseman  172.7  70.31 26.46

1. Which is the oldest team on average? hint: Remember lapply from earlier today and use sapply in a similar fashion. Make use of ?sapply !

tm <- split(mlb_eu_measure, mlb_eu_measure$team)
# sapply returns a vector
which.max(sapply(tm, function(x) mean(x$age)))

## NYM 
##  18

max(sapply(tm, function(x) mean(x$age)))

## [1] 30.52

1. Which position has the youngest players on average?

a <- sapply(pos, function(x) mean(x$age))
a[which.min(a)]

## Starting_Pitcher 
##            28.24

1. What would be a good way to visualize differences in group means by position? E.g.: Age by Position. Is there a position on which players last longer?

nms <- levels(mlb$position)
# let's do a bxoplot
par(mar = c(3, 10, 3, 3))
boxplot(height ~ position, las = 1, data = mlb_eu_measure, horizontal = T)

1. What about size and weight by position?

boxplot(weight ~ position, data = mlb_eu_measure, las = 1, horizontal = T)

boxplot(height ~ position, data = mlb_eu_measure, las = 1, horizontal = T)

!! *10. Write a function that compares the mean height of Starting Pitchers pairwise to another position by running a t-test. Why could such a function by useful? * !!

# truncate level names
levels(mlb_eu_measure$position) <- substr(levels(mlb_eu_measure$position), 1, 
    5)


# define a simple function
compare <- function(var_pos, ref_pos = "Start", data = mlb_eu_measure) {
    ref <- data[data$position %in% ref_pos, "height"]
    var <- data[data$position %in% var_pos, "height"]
    t.test(ref, var)
}

# call the function one time:
compare("Desig")

## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 1.103, df = 21.58, p-value = 0.282
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -1.113  3.639
## sample estimates:
## mean of x mean of y 
##     189.8     188.5

levels(mlb_eu_measure$position)

## [1] "Catch" "Desig" "First" "Outfi" "Relie" "Secon" "Short" "Start" "Third"

# ok let's run a pairwise comparison of starting pitchers with ANY position
lapply(setNames(levels(mlb_eu_measure$position), levels(mlb_eu_measure$position)), 
    compare)

## $Catch
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 7.899, df = 164.9, p-value = 3.761e-13
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  3.802 6.336
## sample estimates:
## mean of x mean of y 
##     189.8     184.7 
## 
## 
## $Desig
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 1.103, df = 21.58, p-value = 0.282
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -1.113  3.639
## sample estimates:
## mean of x mean of y 
##     189.8     188.5 
## 
## 
## $First
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 2.368, df = 93.1, p-value = 0.01994
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  0.2952 3.3597
## sample estimates:
## mean of x mean of y 
##     189.8     188.0 
## 
## 
## $Outfi
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 8.161, df = 412.8, p-value = 4.066e-15
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  3.296 5.387
## sample estimates:
## mean of x mean of y 
##     189.8     185.4 
## 
## 
## $Relie
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 1.762, df = 466.8, p-value = 0.07867
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -0.1008  1.8526
## sample estimates:
## mean of x mean of y 
##     189.8     188.9 
## 
## 
## $Secon
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 12.48, df = 115.7, p-value < 2.2e-16
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  7.175 9.881
## sample estimates:
## mean of x mean of y 
##     189.8     181.3 
## 
## 
## $Short
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 9.66, df = 92.72, p-value = 1.104e-15
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  5.681 8.622
## sample estimates:
## mean of x mean of y 
##     189.8     182.6 
## 
## 
## $Start
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 0, df = 440, p-value = 1
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  -1.069  1.069
## sample estimates:
## mean of x mean of y 
##     189.8     189.8 
## 
## 
## $Third
## 
##  Welch Two Sample t-test
## 
## data:  ref and var
## t = 4.759, df = 65.57, p-value = 1.112e-05
## alternative hypothesis: true difference in means is not equal to 0
## 95 percent confidence interval:
##  2.469 6.040
## sample estimates:
## mean of x mean of y 
##     189.8     185.5