Estimating impact of Predictor

#1
Hi,

I have a simple regression with 24-36 months of data, let's say of Sales which is the dependant variable and % spend on promotions in that month as independent variable (amongst other variables). I am running a normal linear regression on this

The independent variable , let's say can vary from 0-100% and the Dependant variable Sales varies from 7-10K $.

After running the regression, i am getting unusually high beta for the Predictor, 50-60K as the value.

What i want to understand is, is this the right construct of running a linear regression and if any transformation is required? I dont think so, as when i plotted the two on a scatter plot and fit a linear trend like,it fit perfectly and while displaying the R-squared, it showed a value of 60%.

if i want to understand, what is the impact on Sales on a 1% increase in the INdependent variable - how do i do this?

Currently, what i am doing is :

1. Computing Y (Baseline) as (beta * Average X + C)

in my case Beta is 60K, Average X is, let's say 15% and hence i obtain Y (Baseline) as 8.5k

2. For a 1% increase in X, i want to compute Y... so, i keep beta as same, C as same, i add 1% to average X and hence compute Y

i subtract this Y from Y(Baseline) computed in Step #1.

Is there anything i am doing wrong...?

Thanks!
 

hlsmith

Omega Contributor
#2
If this isn't time series the model should be appropriate. The one thing you would need to look out for is, the distribution of IV, and whether a majority of values are at the lower or upper bounds (e.g., 0% or 100%), if that is possible. Meaning can you have 11% or negative percentage.