1
$\begingroup$

I am reading the regression analysis resutls in a paper which showed most of the coefficients b2 and b3 are not significant but the R2 is significant. Eventhough there aren't too many variable in this model(see the model below), that happend. How come?

Is it still necessary to watch b2 and b3 and make conclusion like from b2 we can tell that there is a possitiv correlation between X and Y? The author did that but I am not sure whether one should do that if the coefficient is not significant at all.

enter image description here enter image description here

Improve this question
$\endgroup$
1
  • 1
    $\begingroup$ The existence of an interaction does not necessarily have anything to do with it. When the predictors are correlated, it is quite reasonable to have a significant model overall, without any individual predictors being significant. Read the linked thread for more information. If there's anything you still need to know afterwards, come back here & edit your Q w/ what you've learned & what you still don't understand. $\endgroup$
    – gung - Reinstate Monica
    Commented Jun 14, 2014 at 22:14

1 Answer 1

Reset to default
0
$\begingroup$

Look carefully at the model. The dummy variables are implementing a form of analysis of covariance. The zone of tolerance is determined by the value of the regressor (perceived service).

With that clear, then we can see that the product terms between the dummies and the service score are estimating the change in slope above and below the zone of tolerance.

With this parameterization there is a potential collinearity problem, which is probably the cause of the insignificant partial slopes combined with the overall significant regression.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thank you for the answer. What do you mean by saying "dummy variables are implementing a form of analysis of covariance?" (I am a new comer in SPSS and English is not my mother toe, sorry if everything gets a little bit complicated with me...) $\endgroup$
    – yue86231
    Commented Jun 15, 2014 at 18:15
  • $\begingroup$ Analysis of covariance is to lines as analysis of variance is to points. I said "a form of" because the authors of the study constrained their model to a common intercept. $\endgroup$
    – Dennis
    Commented Jun 16, 2014 at 19:08

Not the answer you're looking for? Browse other questions tagged or ask your own question.