- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
re: Interpreting Coefficients in a Logistic Regression output: what is 1/e^ßX
Posted on 12/6/15 at 9:43 am to TheIndulger
Posted on 12/6/15 at 9:43 am to TheIndulger
quote:manipulating the odds ratio is not required. This is a quest for more knowledge
it's better to try to keep up during the semester than wait til the end
0
Posted on 12/6/15 at 9:49 am to sullivanct19a
quote:What this guy said.
Go to library. Get book from Hosmer and Lemeshow. They are the kings.
And I would add get Paul Allison - Logistic Regression using SAS. This book is the shite in understanding logistic regression with dichotomous data, ordinal data, or non-ordinal data. And you can analyse this with one covariate or several covariates.
The book only costs $49 on amazon.
Are you using R or SAS?
Posted on 12/6/15 at 9:50 am to NoFlexZone
288
Posted on 12/6/15 at 9:54 am to sullivanct19a
quote:that is implied when I scored the cat variable as a 1 or 0. Lol
That variable is dichotomous
quote:that's nitpicking. Fine, one unit increase, since I would be adding new/unseen data to the model.
there is no '1 point increase' in such a variable.
quote:textbook doesn't cover manipulating the odds ratio to better understand stat law. Why? BC I like to ask questions that are beyond what I need to know
Why are you asking these questions? Don't you have a textbook?
Posted on 12/6/15 at 9:54 am to AFtigerFan
quote:thank you good sir, BC this is actually correct
is 1/e^ßX interpreting the odds ratio based on a "0" outcome of the target variable? while e^ßX is for "1" outcome?
Yes
Posted on 12/6/15 at 9:58 am to NoFlexZone
quote:
And I would add get Paul Allison - Logistic Regression using SAS.
Absolutely add this book.
Posted on 12/6/15 at 10:20 am to steeltiger17
quote:
steeltiger17
Yes, nitpicking, because it's clear you don't know what you're talking about. You're not going to realize that unless it's pointed out to you. That's all.
Posted on 12/6/15 at 10:39 am to sullivanct19a
quote:dichotomous is a way to categorize a categorical variable. yea, that's nitpicking
Yes, nitpicking,
1 point increase/ 1 unit increase when interpreting the coefficient is nit picking. that kind of nit picking is saved for a professor who takes off points just to do so. In the real world its understood what it means. sshiiit my professor does it all the time.
quote:
Each coefficient is evaluated in the context of holding all variables at their expected values.
understood by anyone understands regression models. does not needed to be said for someone who understands the question in the OP. unnecessary typing
quote:lol
because it's clear you don't know what you're talking about
This post was edited on 12/6/15 at 11:48 am
Posted on 12/6/15 at 10:44 am to NoFlexZone
quote:
Paul Allison - Logistic Regression using SAS
Another good one is Regression Models for Categorical Dependent Variables Using Stata by Long and Freese. It doesn't goes real in depth as far as theory, but is a good overview of the application of logit regression (among other types of models). Would make a good companion to something like the H&L text.
Posted on 12/6/15 at 10:46 am to UFownstSECsince1950
Damn you smart son!
Posted on 12/6/15 at 10:59 am to steeltiger17
Pi R squared.
No, Pi are round. Cornbread are square.
No, Pi are round. Cornbread are square.
Posted on 12/6/15 at 11:32 am to ldts
quote:the first book i've seen that actually shows interpretation of an outcome not occurring in logistic regression and amazon had this section in the free sample. its done by using 1/e^ßX
Another good one is Regression Models for Categorical Dependent Variables Using Stata by Long and Freese.
This post was edited on 12/6/15 at 11:33 am
Posted on 12/6/15 at 8:46 pm to sullivanct19a
quote:It is still a categorical variable, just with one a binary, or dichotomous, outcome. The number of outcomes does not negate it's categorical status.
That variable is dichotomous, not 'categorical',
quote:And expand on this, the Beta coefficient refers to the change in the outcome variable based on a ONE unit change of that predictor variable, and as you said, all other variables are held constant.
Each coefficient is evaluated in the context of holding all variables at their expected values.
Posted on 12/6/15 at 9:20 pm to steeltiger17
what did you say about my mother??
Posted on 12/6/15 at 9:32 pm to steeltiger17
quote:Just for clarification purposes, your output should have two Sets of results coefficients, one for the constant and one for Mobil. The constant is when Mobil is 0. Just want to make sure you're interpreting the Mobil variable and not the constant if you are talking about the 1 unit increase.
the first book i've seen that actually shows interpretation of an outcome not occurring in logistic regression and amazon had this section in the free sample. its done by using 1/e^ßX
Anyways, if you're looking for other ways to interpret the coefficient, you can determine the probability of the outcome using the odds ratio as well. You may already know this though.
The simplest way is to take the OR/(1+ OR) but you can also get this answer using the beta coefficients, although that can get more complicated. 1/(1+e^-(ß0+b1x1) so if the X is 0 it is essentially 1/(1+e^-(ß0) or 1/(1+e^-(-1.5833)) = .174
Just FYI.
This post was edited on 12/6/15 at 9:56 pm
Posted on 12/6/15 at 9:45 pm to steeltiger17
:wat:
Posted on 12/6/15 at 9:59 pm to steeltiger17
Jesus God, don't you even Fournier Expression brah?
I hope your dental records are up to date.
I hope your dental records are up to date.
Posted on 12/6/15 at 10:12 pm to buckeye_vol
quote:
It is still a categorical variable, just with one a binary, or dichotomous, outcome. The number of outcomes does not negate it's categorical status.
True. But, there's a big difference in categories of color, and saying something is there or it's not. In reading his comments, it was clear he didn't understand that, which is why it needed to be pointed out.
Second part: Evaluate the conditional in the context of marginal expectations. Like how MIT economist Gruber worked to pass the ACA. He counted on the most representative state of the American voter - the moron.
Posted on 12/7/15 at 12:49 pm to soccerfüt
quote:I almost took you seriously with this comment.
don't you even Fournier Expression brah?
What do you know about Fourier Analysis?
Posted on 12/7/15 at 12:51 pm to steeltiger17
Muskogee, Oklahoma
Page 2 of 3
Popular
Back to top
Follow TigerDroppings for LSU Football News