# Thread: Logistic regression diagnostic questions

1. ## Re: Logistic regression diagnostic questions

What software is that function created in (don't think I have seen that form).

I need to know which variables are non-linear which from your comments I don't think you can tell from that.

Amusingly since my variables are commonly binary in nature I don't have to run many of them.

2. ## Re: Logistic regression diagnostic questions

@JesperHP - You need to start math tags without the backslash so you want [math] Math here [/math]

3. ## The Following User Says Thank You to Dason For This Useful Post:

JesperHP (03-20-2013)

4. ## Re: Logistic regression diagnostic questions

Noetsi that sas link was great. I feel like an idiot at times, can someone explain the purpose of finding the non - linearity and what it actually is?

Again, noetsi I appreciate your posts in that you seem to ask many of the questions that I have and usually beat them to death until you better understand them. Unfortunately I have difficulties becoming so unfettered and end up wallowing in my own ignorance. That and you are one of the sas folks around here.

5. ## Re: Logistic regression diagnostic questions

Originally Posted by Dason
@JesperHP - You need to start math tags without the backslash so you want [math] Math here [/math]
thx I forgot how it was done ... tried editing turning slashes the other way didn't work either so I gave up .... figured you would tell me

6. ## Re: Logistic regression diagnostic questions

Originally Posted by hlsmith
Noetsi that sas link was great. I feel like an idiot at times, can someone explain the purpose of finding the non - linearity and what it actually is?

Again, noetsi I appreciate your posts in that you seem to ask many of the questions that I have and usually beat them to death until you better understand them. Unfortunately I have difficulties becoming so unfettered and end up wallowing in my own ignorance. That and you are one of the sas folks around here.
I feel like an idiot all the time in stats

Linearity between the logit and the IV (well the continuous IV) is an assumption of logistic regression. If you can assume linearity then the slope of the IV is true generally at all levels of the IV - that is it does not change. If this is not the case the relationship of the IV to the DV (that is the slope) will be different at different levels of the IV. Note there is a difference between the parameters being linear and the X for that parameter being linear but that is a conversation best left to others. The parameters (slopes) being non-linear is far harder to address.

I beat things to death because I have serious doubts how well I understand much of this and am very afraid of making a mistake. I do this for work and if I make a mistake it could have serious consequence (well if anyone pays attention to what I do).