topic Re: How I interpret the beta coefficient with log outcome in Statistical Procedures

How I interpret the beta coefficient with log outcome

dode — Mon, 13 Nov 2017 13:11:38 GMT

I am doing linear regression when i check my outcome was not normally distributed so I transformed into log and it's worked. Now how I interpret the beta coefficient in log scal?
Is there a way to expandcit it back to interpretation would be easier ?

Re: How I interpret the beta coefficient with log outcome

dode — Mon, 13 Nov 2017 13:11:46 GMT

Re: How I interpret the beta coefficient with log outcome

Ksharp — Mon, 13 Nov 2017 14:47:27 GMT

log(y1)-log(y2)= beta*(x+1 - x)

==>

log(y1/y2)=beta

==>

y1/y2=exp(beta)

it is odds .

Re: How I interpret the beta coefficient with log outcome

PaigeMiller — Mon, 13 Nov 2017 14:56:35 GMT

@dode wrote:
I am doing linear regression when i check my outcome was not normally distributed so I transformed into log and it's worked. Now how I interpret the beta coefficient in log scal?
Is there a way to expandcit it back to interpretation would be easier ?

The response variable does not have to be normally distributed to use linear regression.

The residuals have to be normally distributed to use linear regression, and then the requirement to be normally distributed is used only when doing statistical testing; it is not used when fitting the model.

So it is possible you did not need to perform the log transformation at all.

Re: How I interpret the beta coefficient with log outcome

dode — Tue, 14 Nov 2017 03:27:29 GMT

Yes I checked the residual and it's not normally did so I had to transform my outcome using log. Now how I interpret the beta coefficient?

Re: How I interpret the beta coefficient with log outcome

PaigeMiller — Tue, 14 Nov 2017 18:02:34 GMT

@Ksharp has already explained this.

Wikipedia has even more details.

Re: How I interpret the beta coefficient with log outcome

dode — Tue, 14 Nov 2017 18:32:37 GMT

Sir, I am doing linear regression not logistic regression.