# Multiple Regression Multicollinearity issue

So using the multiple regression model, I am working on finding the relationship between several x factors and the average life expectancy at birth. x1= adult education level (age 25-64) x2= infant mortality rate (per 1000 live birth) x3= CO2 emission rate (megaton) y= average life expectancy

I got the data from OECD statistics and World Bank and conducted a multicollinearity test and the values are just so high and shows multicollinearity. Does this mean I have to change all x factors? I mean the data just shows like this. I mean CO2 emission and Infant Mortality Rate doesn't seem to have any relationship, or not that I know of.... Am I missing some part or do I have to change x factors?

Thanks

• Multicollinearity is not such an issue for predictive modeling (unless it is so strong or perfect that it goofs up floating point arithmetic on a computer). The issue is when you want to do inference on the parameters, as multicollinearity inflates parameter standard errors. What is your interest in building your model.
– Dave
Commented Nov 17, 2020 at 13:32
• There is something very suspicious here as those correlations look much too high for variables like those. What do scattergrams look like? Commented Nov 17, 2020 at 14:31
• So I'm devising this experiment to see whether there is any linear relationships between certain x variables I chose to the average life expectancy at birth in Korea. Yeah I was like waah? when I saw the scattergram as the positive correlation was nearly perfect. I simply googled life expectancy, adult education level, infant mortality rate, and CO2 emission in OECD Stats and World Bank database and found the information in between 1995 and 2016. Maybe there is something I am doing wrong in terms of gathering data?? Commented Nov 17, 2020 at 16:14