ANOVA with a known linear model

I want to use ANOVA function to generate a report for a multiple linear regression model. ANOVA takes the input of LM, a linear model class, which can be created using fitlm function. My question is that since the model is already determined including the coefficients and the predictors, can I input my model somehow in ANOVA without having to run fitlm function? In other words, how can I create a LM class with my own coefficients?

Thanks in advance.

Answer 1

[Ghada Saleh]

Hi Claire,

As you have mentioned, ANOVA only accepts an input of type Linear Model which can be constructed using 'fitlm' or 'stepwiselm'. If you already have the coefficients of your linear model, you can use them to generate data for the dependent variable given your independent variable values. As an example:

x = rand(1,10); % you can replace with your independent variable values.
y = 5*x+12; % replace by your linear model;
lm = fitlm(x,y);

Now 'lm' is a Linear Model object that will estimate the coefficients '5,12'. Given enough number of samples, 'fitlm' will estimate your model and create Linear Model object that you can then input to ANOVA.

I hope this helps,

Ghada

[Brendan Hamm]

Or you could just use the anova1() function, or anova2(), etc.