the coefficients on my spss multiple regression are not significant eventhough my anova is.

I don't know how to explain this but im doing different physical activity intensities and seeing their effect on arterial stiffness. The desired arterial stiffness is low, and those who do more activity have less stiff arteries. But according to my spss activity is not a significant factor, eventhough I know it is and even my scatter graphs show this. I dont know how much sense this is making but I think its back to front if this makes sense.

It's like my correlations too show negative correlation between activity and arterial stiffness eventhough those with high activity have low stiffness and visa versa.

I've attached my spss output I dont think it will let me add my graphs too.

if anybody can help it is GREATLY appreciated.Thanks.enter image description here

  • $\begingroup$ Regarding your correlation, you should expect it to be negative. This is because as your activity increases, your stiffness decreases. That is, the effect works in an opposite direction (hence why the correlation would be negative). $\endgroup$ – Emma Jean Jan 15 '20 at 22:26
  • 1
    $\begingroup$ Also I do see significance in your model, take a look at your p-values once more considering a threshold of 0.05 (psst, time spent sitting). $\endgroup$ – Emma Jean Jan 15 '20 at 22:42
  • $\begingroup$ I am amazed you don't get warnings and error messages, because it looks like the last variable is collinear with the first three. Even if for some reason there isn't perfect collinearity, including all four of them likely makes it impossible to detect individual significance, although as a group they might be significant. $\endgroup$ – whuber Jan 15 '20 at 23:10
  • $\begingroup$ thank you both, as you can probs tell stats isnt my strong point.. $\endgroup$ – pete Jan 15 '20 at 23:15

One of your predictors (Time Spent Sitting per Week) shows a signficance value of .001. The standardized coefficient for your Total Physical Activity predictor is quite high, and would almost certainly show a small significance if it were not highly correlated with other predictors in the model.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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