r - Linear model: comparing predictive power of two different measurement methods -

I am interested in predicting y and studying various two measurement techniques < Code> X1 and X2 . It may be for example that I want to predict the taste of a banana, either by measuring how long it is lying on the table, or by measuring the number of brown spots on banana.

I want to know which one of the measurement techniques is better, should I choose to do only one thing?

I can make a linear model in R:

  m1 = lm (y ~ x1) m2 = lm (y ~ x2)   
 Now we say that the banana taste is better than  X1  this  X2 . When calculating R ^ 2 of two models, the model's R ^ 2 model is clearly more than  m1  model  m2  how  X1  < Better than code> X2 , but before writing a paper, I have some kind of indication that the difference is probably not possible in the form of a P-value  
 How would one go about this? How to do this when I am using different brands of bananas and taking a linear compound effect model which includes the Kena brand as a random effect?   
 
 
 Sorry, if you do not understand right. As far as I can understand, this simple fundamental figure is a question, not R.  
 You put them together in 1 regression. The P-value for each coefficient shows whether they are important or not. You can also brand banana as a banana (if there are not too many types). And ANOVA test are both measurements in BTW different models? What are the R ^ 2 of those models and the combined model? For your problem, look at the definition of R ^ 2, which will help you :)