F value and P value shows and model not fit, but the interaction of indicators is highly significant and there is an increase in the effect. The coefficient variance is high. In this case accept or reject the null hypothesis or there is problem in regression model
How much indicatros You have in the regression? Sample size? If there are to much indicatros You must to do Data rediction first. If two variables are in the same component (factor), that may be a reason for unsignificant regression.
The regression analysis takes for granted that the model is correct. In some cases it is possible that a diagnostic plots of the residuals can reveal striking patterns that may suggest that the model is inappropriate.
If you have no reason to think that the model is inapproriate, then a large p-value tells you that likelihood of your data is not much increased when the model is considered, so the model does not "explain" any particular feature of your data. As this is still a statement about the model and the data, you can not conclude what the reason for this failure (missing explanatory power) is. It is therefore not correct to "accept the null hypothesis", just because the p-value is large.
Dear Waed Alshoubaki,
I suggest to used any feature selection technique to select best indicators for your model. Because some indicators and not good predictor or significant predictor for your model. After selection of best set of predictor, you will use regression analysis to develop your model. I am sure, you will get better result.
The involvement of interaction terms imply that this is a factorial model. The fact that the model is non-significant means that some main effects may not be significant. It is possible in a factorial experiment for main effects to be non-significant but interaction effects significant and the overall model non-significant.
- Badges
- Science topic
- Similar topics
- Chemistry
- Pharmacology
- Pharmaceuticals
- Drugs