OK, we have data collected and normalised, questions are clear (?). Right now lets do some thoughts about statistics in archaeology.

“Statistics theory falls into two camps, frequentist and Bayesian. Frequentist is older and solid. Bayesian is newer, more flexible, and more exciting.”

More about this in a nice George Casella presentation. He mentioned the “likelihoodists” as well but did not concern about them.

frequentists stats

OK, lets play first with this one. Why not descriptive? Because I was sampling and data acquired are only the portion of all data available. Why first? Cause I know nothing about sexy Bayesian… :).

multivariate

Regarding multivariate stats one has to keep in mind:

they are generally divided into **clustering** and **ordinal** types. While clustering groups objects together according similarity in variables, ordinal is arranging them in graph according most valuable (significant) variables

all multivariate methods try to reduce the number of dimensions by searching those most valuable and defining the (hidden) dependencies among them (simplification of view on data)

table of objects with variables (dimensions) is always a starting point

we could analyse relationships **among objects** or **among variables**

every method has its limitation regarding the nature of data (e.g. quantitative, qualitative etc…)

every dataset