I'm dipping my toes into statistics, as usual its choosing a statistical test.

Reading up on tests made me think meta.

Shouldn't all statistical tests performed on anything ever in fact be factorial ANOVA? Sure you can decide to choose one dependent variable and measure it, but how in the hell you're supposed to know if the set of independent variables that you chose actually encompasses the relevancy entirely? Shouldn't all variables about the population be collected and tested exhaustively until it all boils down to a t-test? But you HAVE to collect it all and then you HAVE to boil it down to t-test through factorial ANOVA?

Isn't just using a t-test from the beginning simply improper, because its voluntarism?