LinearRegressionAnalysis
Michael Phillips
3/23/2021
Executive Summary
Aids is a huge predictor of low life expectancy in developing countries, but not at all in developed.
BMI is important in developing countries (lower BMI -> lower life expectancy) but not in developed. Adolescent thinness is associated with lower life in developed countries. This all is evidence that worldwide there is a much bigger problem with undernourishment than with obesity.
Schooling is important indicator for both sets, but extremely important for developing countries.
All Data
First I grouped all data together from the kaggle dataset.
##
## Call:
## lm(formula = Life.expectancy ~ Status + Alcohol + percentage.expenditure +
## BMI + GDP + HIV.AIDS + Diphtheria + thinness.5.9.years +
## Income.composition.of.resources + Schooling, data = df_kaggle)
##
## Residuals:
## Min 1Q Median 3Q Max
## -26.6865 -2.5516 0.0301 2.5593 23.9988
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.885e+01 6.511e-01 75.033 < 2e-16 ***
## StatusDeveloping -1.793e+00 3.221e-01 -5.568 2.89e-08 ***
## Alcohol -1.644e-01 3.041e-02 -5.408 7.03e-08 ***
## percentage.expenditure 1.551e-04 1.066e-04 1.455 0.145738
## BMI 4.863e-02 6.071e-03 8.011 1.80e-15 ***
## GDP 3.395e-05 1.625e-05 2.089 0.036804 *
## HIV.AIDS -6.384e-01 1.688e-02 -37.823 < 2e-16 ***
## Diphtheria 4.770e-02 4.304e-03 11.082 < 2e-16 ***
## thinness.5.9.years -9.047e-02 2.442e-02 -3.705 0.000217 ***
## Income.composition.of.resources 8.967e+00 7.208e-01 12.441 < 2e-16 ***
## Schooling 1.036e+00 5.102e-02 20.300 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4.296 on 2293 degrees of freedom
## (634 observations deleted due to missingness)
## Multiple R-squared: 0.8052, Adjusted R-squared: 0.8044
## F-statistic: 948 on 10 and 2293 DF, p-value: < 2.2e-16
From this it looks like HIV.AIDS, Schooling, and Income Composition are the biggest effects.
Then I wanted to break it up into developed vs developing and see if any different trends.
Developed Countries
##
## Call:
## lm(formula = Life.expectancy ~ Alcohol + thinness..1.19.years +
## Income.composition.of.resources + Schooling, data = df_developed)
##
## Residuals:
## Min 1Q Median 3Q Max
## -4.8293 -1.6025 -0.4617 0.8459 9.8523
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 50.15858 2.94033 17.059 < 2e-16 ***
## Alcohol -0.24222 0.04575 -5.295 1.90e-07 ***
## thinness..1.19.years -1.61176 0.20302 -7.939 1.77e-14 ***
## Income.composition.of.resources 46.44105 3.70112 12.548 < 2e-16 ***
## Schooling -0.37861 0.09522 -3.976 8.21e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 2.543 on 433 degrees of freedom
## (74 observations deleted due to missingness)
## Multiple R-squared: 0.6086, Adjusted R-squared: 0.605
## F-statistic: 168.3 on 4 and 433 DF, p-value: < 2.2e-16
Developing Countries
##
## Call:
## lm(formula = Life.expectancy ~ Alcohol + percentage.expenditure +
## BMI + GDP + HIV.AIDS + Diphtheria + thinness.5.9.years +
## Income.composition.of.resources + Schooling, data = df_developing)
##
## Residuals:
## Min 1Q Median 3Q Max
## -26.7231 -2.5439 0.2099 2.6226 24.2610
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.644e+01 5.505e-01 84.368 < 2e-16 ***
## Alcohol -1.577e-01 3.491e-02 -4.518 6.64e-06 ***
## percentage.expenditure 1.356e-03 2.286e-04 5.933 3.53e-09 ***
## BMI 6.832e-02 7.235e-03 9.444 < 2e-16 ***
## GDP -4.904e-05 2.439e-05 -2.010 0.0445 *
## HIV.AIDS -6.352e-01 1.737e-02 -36.574 < 2e-16 ***
## Diphtheria 4.873e-02 4.650e-03 10.480 < 2e-16 ***
## thinness.5.9.years -3.334e-02 2.577e-02 -1.294 0.1959
## Income.composition.of.resources 7.277e+00 7.548e-01 9.642 < 2e-16 ***
## Schooling 1.064e+00 5.735e-02 18.560 < 2e-16 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 4.402 on 1871 degrees of freedom
## (545 observations deleted due to missingness)
## Multiple R-squared: 0.7713, Adjusted R-squared: 0.7702
## F-statistic: 701.2 on 9 and 1871 DF, p-value: < 2.2e-16
Takeaways
Aids is a huge predictor of low life expectancy in developing countries, but not at all in developed.
BMI is important in developing countries (lower BMI -> lower life expectancy) but not in developed. Adolescent thinness is associated with lower life in developed countries. This all is evidence that worldwide there is a much bigger problem with undernourishment than with obesity.
Schooling is important indicator for both sets, but extremely important for developing countries.