In given Problem, a regression line was fitted using X = life expectancy and Y = crude birth rate. The X‾ = 68.7 years and Y‾ = 21.4 births per 1000 population. Here, the effect of the three types of outliers will be explored.
(a) Add an outlier in Y that has a value (X, Y) of (68.7,44) and recompute the regression line.
(b) Remove the outlier added in (a) and add an outlier in X of (85,11). Recompute the regression line.
(c) Remove the outlier in (b) and add an outlier in X and Y (influential point) of (85,44). Recompute the regression line.
(d) Compare the effects of these three outliers on the slope coefficient and the correlation coefficient.
Problem
The following table contains data from 17 countries. These countries are a systematic sample from twice this number of countries that had a population > 30 million people in 1998 and had complete data. The data were taken from the US. Bureau of the Census, Statistical Abstracts of the United States; 1998, 118th ed., Washington, DC. It consists of the crude birth rate (CBR) or the total number of births divided by the total population in 1998, life expectancy (LifeExp) from birth in years for the overall population, and gross domestic product (GDP) per capita for 1995. The GDP has been adjusted and then converted to U.S. dollars.
Country CBR LifeExp GDP
Argentina 20 74.5 7,909
Brazil 20.9 64.4 4,080
Canada 12.1 79.2 19,000
Colombia 24.9 70.1 2,107
Egypt 27.3 62.1 746
France 11.7 78.5 26,290
India 25.9 62.9 348
Iran 31.4 68.3 2,449
Japan 10.3 80.0 41,160
Mexico 25.5 71.6 2,521
Pakistan 34.4 59.1 482
Poland 9.8 72.8 5,404
S. Africa 26.4 55.7 3,185
Tanzania 40.8 46.4 134
Thailand 16.8 69.0 2,806
United Kindgom 12.0 77.2 19,020
United States 14.4 76.1 27,550
(a) State which model the data in the table follows.
(b) Plot the scatter diagram when X = life expectancy and Y = crude birth rate and fit a regression line.
(c) Does the data appear to follow a straight line?
(d) If life expectancy increased by 10 years, what do you expect to happen to the crude birth rate?
(e) Test that Ho : β = 0. If you test that p = 0, will you get the same P value?