Untitled

[tiredlikedog]

R2 vs R2 adjusted

R2 coefficient (adjusted or plain) says how much of data can be explained by the model.

Hence it means: the higher R2 or R2 adjusted, the better model we have.

R2 can only increase with every added variable.

R2 adjusted can either increase or decrease depending on the value that new variable brings.

We cant say what values of R2 or R2 adjusted are good enough or for what values we should aim as every scientific area and model can reach different values.

Nevertheless, R2 is important statistical measure that helps for comparing between different versions of model during the training process.

[tiredlikedog]

P-Value

p value is probability that coefficient of variable is equal 0.

the higher probability (p-value), the higher chance that coefficient is equal 0 and hence the variable is not important.

the lower probability - the lower p-value, the higher chance the coefficient is important