# Multilinear Regression, test for Dependence?

#### WWGD

Gold Member
Hi All,
Say we conduct a standard linear regression test of Y (dep) versus X (independent)

Then there are tests to be made on whether there is a linear relationship between Y and X

(with $H_o$ being that m=0; m is the regression line slope versus $H_A :m \neq 0$)

Is there a similar test for multilinear regression, to determine linear dependence

of Y versus $X_1, X_2,..,X_n$ ?

Thanks.

Last edited:
Related Set Theory, Logic, Probability, Statistics News on Phys.org

#### WWGD

Gold Member
Danke.

• fresh_42

#### FactChecker

Gold Member
2018 Award
Yes. There is a lot available for multiple linear regression. You can test if Y is statistically the entire set of independent variables. You can also use a "stepwise" linear regression algorithm that will only end up with a set / subset of independent variables that are statistically needed for the regression. The statistical software package R is free and has good regression algorithms. ( See stepAIC in http://www.statmethods.net/stats/regression.html ). The algorithm will give you p-values that tell you the statistical significance of the model and the individual variables.

• WWGD

#### WWGD

Gold Member
Yes. There is a lot available for multiple linear regression. You can test if Y is statistically the entire set of independent variables. You can also use a "stepwise" linear regression algorithm that will only end up with a set / subset of independent variables that are statistically needed for the regression. The statistical software package R is free and has good regression algorithms. ( See stepAIC in http://www.statmethods.net/stats/regression.html ). The algorithm will give you p-values that tell you the statistical significance of the model and the individual variables.
Thanks, do you think PCA would be in order here, to determine which of the $X_i$ have more weight in determining the value of $Y$?

#### FactChecker

Gold Member
2018 Award
Thanks, do you think PCA would be in order here, to determine which of the $X_i$ have more weight in determining the value of $Y$?
No. PCA has a different use. It tries to represent the spread of data the best using fewer dimensions. But it does not single out one variable, Y, to explain, estimate, or predict. In fact, it might give you a linear combination that is very bad at estimating Y. If you want to find the best model for estimating Y = f(X), f linear, then you should use linear regression.

#### WWGD

Gold Member
Yes. There is a lot available for multiple linear regression. You can test if Y is statistically the entire set of independent variables. You can also use a "stepwise" linear regression algorithm that will only end up with a set / subset of independent variables that are statistically needed for the regression. The statistical software package R is free and has good regression algorithms. ( See stepAIC in http://www.statmethods.net/stats/regression.html ). The algorithm will give you p-values that tell you the statistical significance of the model and the individual variables.
Sorry to bother, but I can't find the step AIC. Would you please help?

Gold Member
2018 Award

### Physics Forums Values

We Value Quality
• Topics based on mainstream science
• Proper English grammar and spelling
We Value Civility
• Positive and compassionate attitudes
• Patience while debating
We Value Productivity
• Disciplined to remain on-topic
• Recognition of own weaknesses
• Solo and co-op problem solving