search

ijyliu / ECMA-31330-Project

Econometrics and Machine Learning Group Project
2 stars 1 forks source link

Correlation between covariates #36

Closed ijyliu closed 3 years ago

ijyliu commented 3 years ago

https://github.com/ijyliu/ECMA-31330-Project/blob/main/Output/Figures/LE_Health_Econ_Correlations.pdf

It is strong only between groups of covariates, for example the GDP measures, where there is a big white block.

GDP itself appears to be the most strongly related to life expectancy, the other items seem kind of irrelevant.

ijyliu commented 3 years ago

The thing is, earlier when I was using just GDP related measures (like on Tuesday), one would think the 'rho' in @marionoro 's simulations would be higher, since those items are correlated. But even then the results for a difference between OLS and including the PC1 were not large

ijyliu commented 3 years ago

I guess I should invert the poverty measures. Though the correlation with GDP, etc. is not strong in magnitude regardless...

ijyliu commented 3 years ago

Currently using just the GDP measures, among which the correlation is sufficiently strong (above 0.5 I think). The rho correlation with the variable of interest is in a similar range.

ODA and poverty measures have been inverted, but probably won't be used.