I performed a LR model and I know that only the p-value and RR^2 are not enough. Any help?
Since you discuss accuracy I will presume prediction is your goal. Therefore I recommend the attached screenshot book available for download from the z-library Best wishes David Booth
Dan Li, Check VIF. Select your variables using the values of t. There are many other measures to look at for model accuracy. AIC, BIC and S measures actually reflect the fitment of models. More important is what the business needs from the model. It’s not always the statistical inferences but business inferences are equally important
Dan Li lowest AIC as you run more and more iterations. Test and train variance should not exceed 10% for RMSE.
Dan Li , you say "I know that only the p-value and RR^2 are not enough". What else do you want to know ?
You might want to look at measures like Mean absolute error (MAE), Mean absolute percent error (MAPE), or Normalized root mean square error (NRMSE). These are sometimes considered measures of "accuracy" because they compare the values predicted by the model to the actual values of the dependent variable.
( But also note that for ordinary least square models (OLS), the R-squared can also be interpreted as comparing predicted to actual values. )
In any case, it may be helpful to plot the predicted values vs. the actual values to see if the model is consistent in being "accurate" across the range of observed values.
- Badges
- Science method