Social Epidemiology
So what does this all have to do with social epidemiology?
PCA has been used in social epi in a number of different ways and it gets much more useful when there are multiple dimensions of data. PCA is used to explore relationships between variables and relationships of variables to particular clusters in the dataset. It is good for elucidating what particular constructs might be underlying the data.
Sometimes, we can use the first principal component (or first and second principal components - they are uncorrelated) as a proxy measure of some underlying construct for the variables. If you had data on several measures of socioeconomic position, for example, it would be possible to use PC1 as a proxy variable for the underlying construct of SEP and to thus adjust for it in an analysis. This is what we will do in the PCA practicals.
1.6 Social Epidemiology
So what does this all have to do with social epidemiology?
PCA has been used in social epi in a number of different ways and it gets much more useful when there are multiple dimensions of data. PCA is used to explore relationships between variables and relationships of variables to particular clusters in the dataset. It is good for elucidating what particular constructs might be underlying the data.
Sometimes, we can use the first principal component (or first and second principal components - they are uncorrelated) as a proxy measure of some underlying construct for the variables. If you had data on several measures of socioeconomic position, for example, it would be possible to use PC1 as a proxy variable for the underlying construct of SEP and to thus adjust for it in an analysis. This is what we will do in the PCA practicals.