Exercises: Week 4

Data and Computing Fundamentals

Data Cleaning

Implement the steps for cleaning the OrdwayBirdsOrig data, including

• Cleaning the species name using the crowd-sourced correction file. (This is done in the write-up.)
• Cleaning the year, converting to proper numerical form. (This is done in the write-up.)
• Clean the month. Convert it to a valid number: 1 through 12
• Clean the day. Convert it to a valid number: 1 through 31.
• Clean the weight. You can use a strategy similar to that used for WingChord.
• See which Data Entry Person makes more mistakes in entering the SpeciesName. Hint: compare SpeciesName to SpeciesNameCleaned.

MistakeSubset = subset(namesCleaned, SpeciesNameCleaned = !SpeciesName)

## Error: object 'namesCleaned' not found

MistakeSubset

## Error: object 'MistakeSubset' not found

Using the Cleaned Data

• Make a scatter plot of wing chord versus weight for different species.

I have no idea how to approach this problem.

Using Grouped Data for Individual Cases

Using the Cleaned Ordway Bird data:

• Construct a subset of the bird data consisting of those bird species that have more than 200 cases altogether.
• Construct a subset of bird data where there are more than 200 cases altogether or more than 20 in any month of the year.
• Construct a subset of the bird data consisting of those months and years where there are more than 50 birds in the record and the species is more than 5% of all the birds in that year.

Country Data

The countrySynonyms data file (you can load it with data(countrySynonmyms)) gives word synonyms for each country listed in the World-Map software, together with an official ISO 3-letter country code. This data set is in “wide” format. To turn it into narrow format, you can do this:

data(countrySynonyms)
foo <- melt(countrySynonyms, id.vars = c("ID", "ISO3"), value.name = "Country", 
    measure.vars = names(countrySynonyms)[-(1:2)], variable.name = "whence")
countrySynonymsLong <- subset(foo, !is.na(Country))
countrySynonymsLong$whence <- NULL

• Using this countrySynonymsLong, find the names of any countries in the Gapminder data that don't have corresponding names in countrySynonymsLong. Hint: join() and subset() are good tools.
• As best you can, for those countries that don't have an exact match, find the country in countrySynonymsLong that is the non-exact match to the name and fix the Gapminder data accordingly, so that every possible country in the Gapminder data shows up in the map.

The countryRegions data file (you can load it with data(countryRegions) or get documentation with help(countryRegions) let's you aggregate countries in various ways.

• Make a map showing which countries have per capita GDP above and below the average for that continent.

Model Fitting

Height versus Age

data(nhanes)
nhanes = transform(nhanes, sex = ifelse(sex == 2, "F", "M"))
small = sample(nhanes, size = 250)
ggplot(data = small, aes(x = age, y = hgt)) + geom_point() + aes(colour = sex)

## Warning: Removed 41 rows containing missing values (geom_point).

• Compare boys and girls aged 10 years and less.

tenAndUnder = subset(small, age <= 10)
ggplot(data = tenAndUnder, aes(x = age, y = hgt)) + geom_point() + aes(colour = sex)

## Warning: Removed 30 rows containing missing values (geom_point).

mScatter(tenAndUnder)

## Loading required package: manipulate

## Warning: there is no package called 'manipulate'

## Error: could not find function "error"

* Do the data given any reason to think that boys and girls grow differently up to age 10.
early on it seems that boys might be taller, but the difference if there actually is any is mostly gone if not reversed by eight or so.

* Is there any evidence for nonlinear growth in height for these kids?
The graph looks pretty linear to me, but it is possible that it isn't. It 

might just be a few outliers, but it seems that the rate of growth is increasing by the age of ten. It might be helpful to examine more of the graph.

• Looking at boys between 10 and 25 …
  • Make a graph that gives a compelling view of when boys' heights level off with age.

male25andLess = subset(small, age <= 25, sex = "M")
ggplot(data = male25andLess, aes(x = age, y = hgt)) + geom_point() + aes(colour = sex)

## Warning: Removed 32 rows containing missing values (geom_point).

Further Exploration:

Height

Look at people aged greater than 20 years, first with a very small sample (about 100 people) and then with the entire data set.

MoreThan20Small = sample(nhanes, size = 100)
ggplot(data = MoreThan20Small, aes(x = age, y = hgt)) + geom_point() + aes(colour = sex)

## Warning: Removed 21 rows containing missing values (geom_point).

ggplot(data = nhanes, aes(x = age, y = hgt)) + geom_point() + aes(colour = sex)

## Warning: Removed 4609 rows containing missing values (geom_point).

• Does the small data set clearly show a decrease in height with age? How about the entire data?
  The small set debateably shows a decrease in height with age (it might just be due to some outliers), but the entire data makes the trend very clear.

• Any evidence for nonlinear trends in height with age for those greater than 20 years old?
  Here again it might just be outliers, but it seems that the rate of height decrease is increasing at the point that the data ends. It is questionable though, and since shrinking is allegedly not as big an effect as I thought, I cannot think of a very good reason for this to be the case.

BMI

• Do boys and girls show a systematically changing BMI with age? (Hint: Look to see if the confidence interval is consistent with a level-line model.)

ggplot(data = nhanes, aes(x = age, y = bmi)) + geom_point() + aes(colour = sex)

## Warning: Removed 4967 rows containing missing values (geom_point).

• Are the boys and girls different? If not, at what age do males and females diverge in BMI?

For Math 135 graduates …

The BMI models weight as being proportional to height squared. This should give a straight line on a graph of log-weight against log-height. Is this a reasonable model for adults?