### Regression not speaking business well !

Often during Linear Regression modeling, we come across a negative intercept and it becomes quite difficult for us to explain the business sense of the same.Suppose equation comes like : Y = -100 + 23 (Media spend) + 13 (Discount) ; in this case, client argues that the equation means that Sales would be negative if there is no media spend and also discount is nil.

Doesn't make any sense, right ? How can we deal with such situation/equation ? Let me give you a tablet that can treat the model.

### What is Intercept ?

In the equation of simple linear regression >>

**Y = mX + C**,**C is intercept. The only explanation that I consider to be perfect is " The value of Y @ X = 0".**
But what is X can't be zero, isn't the C unrealistic ? Yes, it is !

We generally don't bother about this in the modeling exercise as clients don't bother about equation. But these days people have got smart. So better be prepared.

### Suggest me something !

In the situations, while your X can't be zero and your intercept is coming negative and you want to make it positive, Let me prescribe something ...

RX.

RX.

Instead of modeling Y on X, model it on X - Mean(X) ... so what would the new equation?

###
**Y = m (X - mean(X)) + C**

Now can you tell me, what is C ? Well, it is value of Y at X = mean(X), which is much more realistic scenario than X being 0. Also this trick ( with iterations) would help you treat the intercept to a positive horizon.

Enjoy reading our other articles and stay tuned with ...

Kindly do provide your feedback in the 'Comments' Section and share as much as possible.

### FYI !

Your model was not wrong while intercept was negative, it was only not much acceptable by business sense.Enjoy reading our other articles and stay tuned with ...

Kindly do provide your feedback in the 'Comments' Section and share as much as possible.

I enjoyed this article and it provided me a new solution for problems

ReplyDeleteThanks that was very helpful. However, how do you interpret the coefficients in the regression output. Thank you

ReplyDelete