Significant Correlations - No Sig Beta Weight

#1
Please help me with Multiple Rgression Analysis

Hi guys,

I'm hoping that someone can clear up my confused state of mind...:(

I have conducted a standard MRA and my problem is that even though all of the correlations in the matrix seem good, the CI's for all of the explanatory variables (except V3) encompass positive, negative value and zero and are non significant (p . > .05).

The regression coefficient for V1 was β = -.19 (95% CI = -.71 – .04);
V2 was β = -.09 (95% CI = -.99 – .39)
V3 was β = .28 (95% CI = .19 – 1.04)
V4 was β = .18 (95% CI = -.005 – 1.03)
V5 was β = -.15 (95% CI = -.58 – .11)

The confidence limit for the ‘V3’ variable was the only one that did not encompass both zero and a negative value. This indicates that V3 (β = .28, p < .05) demonstrated significant effects on the perceived scores.

Does this mean that 'V3' is the only significant predictor of the DV? That is the way that I have written up the discussion section and have rejected the hypothesis.

Could someone please help to confirm if those 4 explanatory variables (V1, V2, V4 and V5) are non significant.

Kind regards,
Surfer.
 
Last edited:
#2
It means that only V3 is a predictor of the DV when controlling for the other IVs. Which is not the same as saying that none of the other IVs are predictors of the DV, period.

Let me give an illustrative example. Suppose you are examining the effect of age and arterial plaque on heart attacks. You measure age, arterial plaque, and heart attacks in a large group of people. As you woudl expect, there are significant positive correlations between age and heart attacks and between arterial plaque and heart attacks when each are compared individually. But it's quite possible that, when you do a multi-variate fit, you find that only arterial plaque qualifies as a predictor of heart attack.

What's going on in this hypothetical example is that arterial plaque is the real cause of heart attacks, so once you control for it, age is not a predictive factor. But that does not mean that age isn't a predictor of heart atacks, period. It is, because older people are likely to have more arterial plaque. If you didn't know someone's arterial plaque level, then knowing they are older would skill be a useful clue that they are more likely to have a heart attack. But if you did know that person's arterial plaque level, you should just use it and ignore his age when predicting how likely he is to suffer a heart attack.

Such entirely consistent, but somewhat counter-intuitive, states of affairs go by the fancy name of "multi-colinearlity effects."
 
#3
Hi ichbin,

I think that if you were my Statistics lecturer, I would learn to enjoy the subject alot more. Thankyou very much for the very informative reply! Your illistrative example really helped to put the situation into perspective.

I have been struggling with this MRA for some time, cheers!

Kind regards,
Surfer