PCA and times series

[Cyril Fouillet]

Dear Members of Stata Forum,

We are two economists interested in PCA applications.

Our objective is to build an index of financial inclusion using 6 variables in time series (1969-2015).

We use PCA in order to determine the weights of these variables in the index.

The economic literature often uses “simple” PCA even with time series in order to determine the weights of these variables.

We are aware that this procedure can be misleading because of the non-independence of the observations in case of time series.

So, our question is: Since we are using PCA only to determine the weights (through normalization of components’ loadings), should we abandon simple PCA (and use MSSA) or is it fine?

Thank you very much for your help and time.

[Phil Bromiley]

You'll have a greater chance of a useful answer if you follow the FAQ on asking questions -- provide Stata code in code delimiters, Stata output, and sample data using dataex.

This is really a substantive rather than a Stata question. You might get an answer but it is hit or miss.

You can do a wide variety of PCA-like estimators in Stata both through things included in the multivariate group of routines and the GSEM set of routines.