Quote:
Originally Posted by RonCo
By definition, any piece of data that is too small to determine the root of the outcome means you are unable to determine what skills are in place in the model. Assuming a 1-week sim, you would be unable to determine if, say, Power, was in the model. So, sure, it's easy to create a model in which small sample sizes hide an impact.
I think that's different from your original statement.
Or, if it's not different from your original statement then your original statement was so elementary that it's not particularly valuable in the conversation to my way of seeing things. Shrug.
|
You have two models, M1 and M2 that have important differences. You want to determine which model generated the data you have. But if M1 and M2 generate very similar data, the data can't speak about which model generated it. This is an important statistical point, and directly relevant to the existence of clutch hitting in the real world.
I've said this a couple of times. But it seems you are interested in other things. That's okay. People are interested in different things.