Data Dive 14

library(tidyverse)

## ── Attaching core tidyverse packages ──────────────────────── tidyverse 2.0.0 ──
## ✔ dplyr     1.1.4     ✔ readr     2.1.6
## ✔ forcats   1.0.1     ✔ stringr   1.6.0
## ✔ ggplot2   4.0.1     ✔ tibble    3.3.1
## ✔ lubridate 1.9.4     ✔ tidyr     1.3.2
## ✔ purrr     1.2.1     
## ── Conflicts ────────────────────────────────────────── tidyverse_conflicts() ──
## ✖ dplyr::filter() masks stats::filter()
## ✖ dplyr::lag()    masks stats::lag()
## ℹ Use the conflicted package (<http://conflicted.r-lib.org/>) to force all conflicts to become errors

library(gapminder)
library(ggplot2)
gapminder_raw <- gapminder_unfiltered
gapminder <- gapminder_unfiltered |>
  rename(life_exp = "lifeExp",
         population = "pop",
         gdp_per_cap = "gdpPercap") |>
  mutate(years_since = year(now()) - year)

Goal 1: Business Scenario

Customer or Audience:
- We represent a non-profit taking the place of USAID and we are looking for where our services would do the most good. We only have funding, time, and personnel to place ourselves into three countries to start, thanks to a grant we received from a wealthy benefactor. Depending on our work, we have the opportunity to extend the grant, which we will be able to use to expand our operations and move towards helping more people.
Problem Statement:
- Which three countries would allow us to have the highest year per dollar return on our investment in the first five years of our grant?
Scope: What variables from the data (in the lab) can address the issue presented in your problem statement? What analyses would you use? You’ll need to define any assumptions you feel need to be made before you move forward.
- Our response variable will be life expectancy, and our target variables will be country and gdp per capita.
- Linear Regression, with a hint of weighted least square means.
- Assumptions
  - GDP per capita is related to life expectancy.
  - Life expectancy is a good measure of where people need additional resources and there is an opportunity to help.
  - All else being equal, the country specifically does not have an effect on life expectancy.
  - All else being equal, the continent the people exist on does have an impact on their life expectancy.
  - Life expectancy is positively correlated with year.
  - The United Kingdom is a valid representative country in infant mortality for a developed country.
  - Afganistan is a valid representative country in infant mortality for an under-developed country.
Objective:
- Identify which variables help determine which countries have the lowest life expectancies so that we can determine which countries would therefore benefit from our charity’s services and support.

Goal 2: Model Critique

Improvement 1

The data uses mean life expectancy which can be heavily skewed by higher infant mortality rates. If you look below, you can see the difference in life expectancy over time for a country with high infant mortality (Afganistan) and low infant mortality (United Kingdom). We propose adding a variable for infant mortality rate to help control for this effect

    gapminder$infant <- ifelse(gapminder$country == "United Kingdom", 3.9, NA)
    gapminder$infant <- ifelse(gapminder$country == "Afghanistan", 50, gapminder$infant)

    gapminder_filter <- gapminder %>%
      filter(country %in% c("United Kingdom", "Afghanistan"))

    ggplot(data = gapminder_filter, aes(x = year, y = life_exp)) + 
      geom_point(aes(color = country)) + 
       labs(x = "Year",
                y = "Life Expectancy",
                title = "Life Expectancy for Countries with Low and High Infant Mortality")

Improvement 2

Additionally, we suggest using a generalized linear model instead of the bootstrapping. This will help determine the effect of each independent variable, so we can better make decisions in the future. We use a Poisson glm to account for the fact that the life expectancy variable more closely approximates a Poisson distribution.

model <- glm(round(life_exp) ~ year + gdp_per_cap + continent, data = gapminder, family = "poisson")
summary(model)

## 
## Call:
## glm(formula = round(life_exp) ~ year + gdp_per_cap + continent, 
##     family = "poisson", data = gapminder)
## 
## Coefficients:
##                     Estimate Std. Error z value Pr(>|z|)    
## (Intercept)       -2.458e+00  2.768e-01  -8.882   <2e-16 ***
## year               3.203e-03  1.398e-04  22.913   <2e-16 ***
## gdp_per_cap        3.675e-06  2.157e-07  17.038   <2e-16 ***
## continentAmericas  2.785e-01  8.210e-03  33.919   <2e-16 ***
## continentAsia      2.076e-01  7.960e-03  26.077   <2e-16 ***
## continentEurope    3.393e-01  7.279e-03  46.617   <2e-16 ***
## continentFSU       2.794e-01  1.180e-02  23.666   <2e-16 ***
## continentOceania   3.006e-01  1.073e-02  28.012   <2e-16 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## (Dispersion parameter for poisson family taken to be 1)
## 
##     Null deviance: 7605.1  on 3312  degrees of freedom
## Residual deviance: 2260.4  on 3305  degrees of freedom
## AIC: 22152
## 
## Number of Fisher Scoring iterations: 4

Improvement 3

Finally, the data is imbalanced when it comes to records for individual countries. European countries have records for more years than countries in Africa or Asia. This means the the European data is over represented in the data. We would either need to collect data from the other countries to balance the data set, or condence into one record per decade to better balance the data.

gap_group <- gapminder %>%
  group_by(country) %>%
  summarise(n_records = n(), 
            continent = continent) %>%
  distinct()

## Warning: Returning more (or less) than 1 row per `summarise()` group was deprecated in
## dplyr 1.1.0.
## ℹ Please use `reframe()` instead.
## ℹ When switching from `summarise()` to `reframe()`, remember that `reframe()`
##   always returns an ungrouped data frame and adjust accordingly.
## Call `lifecycle::last_lifecycle_warnings()` to see where this warning was
## generated.

## `summarise()` has grouped output by 'country'. You can override using the
## `.groups` argument.

ggplot(data = gap_group, aes(x = reorder(country, n_records), y = n_records)) + 
  geom_point(aes(color = continent)) + 
  theme(axis.text.x=element_blank()) + 
   labs(x = "",
            y = "Survey Records",
            title = "Number of Observations by Continent")

Goal 3: Ethical and Epistemological Concerns

Overcoming biases (existing or potential).
- There is no information in the data set that details if a country wad involved in a (civil) war, epidemic, or other mass casualty event that skews the median life expectancy.
- There is an assumption that all countries in the data set are universally recognized with no areas in dispute or under occupation.
- There may be an overrepresentation in the data of people living in urban areas in lieu of those living in rural areas due to ease of collecting data.
Possible risks or societal implications.
- If you do the analysis wrong, you could get an underwhelming or under expectations result based on potential investment.
Crucial issues which might not be measurable.
- Quality of life.
Who would be affected by this project, and how does that affect your critique?
- Donors and philanthropic individuals who seek to gain the most well-being per dollar of donation who rely on analyses like this to make informed decisions when making a donation.