If one wants to deal strictly with data instead of a reanalysis product, then stay away from MEI.

>"The Multivariate ENSO Index (MEI) on the six main observed variables over the tropical Pacific. These six variables are: sea-level pressure (P), zonal (U) and meridional (V) components of the surface wind, sea surface temperature (S), surface air temperature (A), and total cloudiness fraction of the sky (C). The MEI is computed separately for each of twelve sliding bi-monthly seasons (Dec/Jan, Jan/Feb,..., Nov/Dec). After spatially filtering the individual fields into clusters (Wolter, 1987) , the MEI is calculated as the first unrotated Principal Component (PC) of all six observed fields combined. This is accomplished by normalizing the total variance of each field first, and then performing the extraction of the first PC on the co-variance matrix of the combined fields (Wolter and Timlin, 1993). In order to keep the MEI comparable, all seasonal values are standardized with respect to each season and to the 1950-93 reference period."

If we are really trying to get to the bottom of mechanisms for ENSO, why would we use a dataset that has gone through so many contortions as this?

So there are two issues with MEI -- (1) the PC deconstruction and (2) reanalysis to get at values that were never even measured, i.e. surface wind from years ago.