Multi-regression | How to rank the weight/ contribution of each variable?

[LIVIA]

Let's think on sales: I have several variables that can drives sales.

I already did a multi regression analysis to understand the variables/ components that drives sales (via LINEST formula).

Now I want to understand the weight of each variable.
Does anyone knows how to RANK the "importance" of each variable? I want to understand which is the most important variable from the regressions, to the least important.

Who knows that??

THANKS!

[zbor]

Can you upload example?

[LIVIA]

Hi,
Here is attached. I tried to explain what is my question again via step by step.
I only need to know how do I understand the right weight of each variable from the multi regression model, since each variable is in a different scale and can not be comparable directly via the coefficient.

Tks,

[zbor]

I can't see formula. Most important value is value with biggest potention.

so, in Ax^2 + Bx + C -> Ax^2 determines formula.

[LIVIA]

Right, I'm not looking necessarily to a formula.
It is more a statistical question: I only need to know how to understand the real WEIGHT of each variable.

BUT, I think it is not only the `biggest value´, since each variable is in a different scale.

It would work, if all variables were in the same conversion, for instance "%". But in this case, each variable has a different scale (Price index vs. competitor, %of promotions, etc). So for this reason, I think I cant only look for the higher constant.

Any idea?!!
THANKSSSS

[shg]

How about normalizoing the independent variables and taking the partial derivative of the regression with respect to each?

[LIVIA]

humm, How do I do that?!! (Im reading something about calculate the T-value of each variable)...

HELP!!!

[shg]

Help for LINEST descibes computing t statistics; it's the slope divided by the standard error for each variable. I don't know whether the absolute magnitude differentiates the 'most important' variable.