1
$\begingroup$

I got the equation for logistic reg, and I am comfortable with the result. Let's say logit(p), ln(p/q), or the model is something like
$$\text{logit}(p) = b+a_1X_1 + a_2X_2 + a_3X_3$$ For example --> $b = 10 , a_1 = 0.5 , a_2 = 0.6 , a_3 =0.7$

So my equation is $\text{logit(p)} = 10 + 0.5X_1 + 0.6X_2 + 0.7X_3

Let's say if I want my probability to be 40% ( $p = 0.4$ ) to be my cut off rate in the system.

How can I use this in the live system?

I'm thinking of giving the developer this logit equation but use in the live system this equation and block account that have fall below success rate threshold?

$\endgroup$
  • 2
    $\begingroup$ Could you please elaborate on what specifically it means to "use" a regression model in a "live system"? $\endgroup$ – whuber Jan 18 '13 at 21:54
1
$\begingroup$

Here is how I would approach it:

Do the logistic regression in R, make the data into an R object, The estimated regression is then an object, and prediction can be done using it via the predict() function. Put this data object and regression model object into an R package. Then the developer programming the live system need only load this package!

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.