Hacker News new | ask | show | jobs
by fsckboy 534 days ago
the goal is for them to be independent variables in the statistical sense. science is a process of refinement so there will undoubtedly be improvements, but they are as independent as they can be based on current knowledge, they can be measured separately, and people exist in every combination (the scales of each measure are not dependent on one another), and we can't describe what we know about personality without including all of them. Myers-Briggs comes to mind as a comparable metric which has more dependence among its variables, and essentially ignores neuroticism.
1 comments
thaumasiotes 534 days ago
> the goal is for them to be independent variables in the statistical sense.

No, it isn't. The goal is for them to have high explanatory power. They aren't independent variables, and the fact that they aren't is a frequent criticism of them. But it's not relevant to whether they are useful descriptors or predictors.

Compare https://en.wikipedia.org/wiki/Big_Five_personality_traits#Li... :

> In many studies, the five factors are not fully orthogonal to one another; that is, the five factors are not independent. Orthogonality is viewed as desirable by some researchers because it minimizes redundancy between the dimensions.

link
fsckboy 534 days ago
that description does not disagree with me.

It was developed by people fully aware of what orthogonality is, and understanding of why orthogonality is desirable. The fact that it is not perfect is indicative of the complexity of psychology and the bluntness of our tools. There is no system that is more orthogonal that explains near as much (I qualify it that way because dead vs alive is simple and pure but of little explanatory power).

We have nothing better than the Big Five (there are proposed refinements but they are not yet generally accepted), and it is better than what came before. The Big Five separates variables that used to be mixed, and can't yet explain the unexplained, but neither can any other system.

link
thaumasiotes 534 days ago
> that description does not disagree with me.

Unless you think that "these five variables appear to be independent when measured" conflicts with "these five variables are not independent".

It doesn't seem like that tricky of a question, but maybe I'm missing something.

link