mdelrosa / data-science-work Goto Github PK

• [1] I love the idea of "Epistemic status"; I may actually require something like this in the future....

• [2] Note that you can add color labels by assigning a variable in aes(); for instance replace this:

gapminder %>%
  filter(year == year_min | year == year_max) %>%
  ggplot(aes(x = continent, y = gdpPercap)) +
  geom_boxplot() +
  geom_point(data = outliers_all, aes(x = continent, y = gdpPercap), color = 'red')+
  facet_wrap(~ year)

With the following for labels:

gapminder %>%
  filter(year == year_min | year == year_max) %>%
  ggplot(aes(x = continent, y = gdpPercap)) +
  geom_boxplot() +
  geom_point(data = outliers_all, aes(x = continent, y = gdpPercap, color = country))+
  facet_wrap(~ year)

• [1] With your q1 vis, you could have investigated which bacteria are treatable with which bacteria with greater resolution than what you present in your observations.

• [2] Also with your q1 vis, some of your points overlap; you could fix this with a position = "dodge" in your geom_point().

• [3] Seems like in q2 you were still focused on whether MIC < 0.1, rather than the prompt of the question.

Comments

[1] Nice job on q1!

[2] Nice job on generalizing the calculation for q4!

[3] We haven't covered this yet, but when you have multiple factors that you're trying to visualize, putting the factors with more levels along coordinate axes is often more effective. Compare:

as_tibble(Titanic) %>%
  filter(Survived == "Yes") %>%
  ggplot(aes(x = Sex, y = n, fill = Class)) +
  geom_bar(stat = "identity", position = position_dodge())

with

as_tibble(Titanic) %>%
  filter(Survived == "Yes") %>%
  ggplot(aes(x = Class, y = n, fill = Sex)) +
  geom_bar(stat = "identity", position = position_dodge())

[4] Also, know that geom_col() is basically a shortcut for geom_bar(stat = "identity").

[5] Your code for q5 would be a lot more succinct with a facet_grid( ~ Sex), rather than the manual-composite plot you've built up, for instance:

df_prop %>%
  filter(Survived == "Yes") %>%
  ggplot(aes(x = Age, y = Prop, fill = Class)) +
  geom_bar(stat = "identity", position = position_dodge()) +
  scale_y_continuous(breaks=c(0.0, 0.25, 0.5, 0.75, 1.0)) +
  facet_grid( ~ Sex) +
  ylim(0, 1.05)

Comments

[1] I disagree with the assertion "cut does not seem to mediate price." Rather I'd say "cut does not seem to significantly mediate price, as compared with carat".\0

Comments

[1] I don't understand why you got rid of the knitr::kable() call... it's really great for making pretty markdown tables....

[2] I'm not really sure what you're doing with this ddply call:

mu <- df_err %>%
  ddply(
        "Distinctness", summarise,
        grp.mean = mean(VelocityError),
        grp.sd = sd(VelocityError),
        grp.abs_mean = mean(abs(VelocityError)),
        grp.abs_sd = sd(abs(VelocityError))
        )
mu

We already learned how to use group_by() to do a similar operation

df_err %>%
  group_by(Distinctness) %>%
  summarize(
    grp.mean = mean(VelocityError),
    grp.sd = sd(VelocityError),
    grp.abs_mean = mean(abs(VelocityError)),
    grp.abs_sd = sd(abs(VelocityError))
  )

Personally, I find summaries-grouped to be a lot more readable.

[3] Excellent work posing and testing hypotheses!

Comments

• [1] For q3 you state "Both properties decrease with increasing thickness, but not linearly." This is slippery: while both properties appear to decrease, we can't conclusively say that they do decrease.

Furthermore we should generally distrust the edges of a regression, as that's where our fit is least trustworthy.