I am trying to work out what precisely this means. Does the points being tightly packed around the middle of the curve just mean that my model isn't that good? Here is my GLM call:

logitmodel=(glm(status~mhi+mgr+edu,data=Rust_Data_10_13,family = binomial)), the summary looks like this:

glm(formula = status ~ mhi + mgr + edu, family = binomial, data = Rust_Data_10_13)

Deviance Residuals: 
    Min       1Q   Median       3Q      Max  
-1.2804  -0.7331  -0.6343  -0.5115   2.0189  

              Estimate Std. Error z value Pr(>|z|)    
(Intercept) -2.072e+00  3.733e-01  -5.551 2.84e-08 ***
mhi         -8.225e-06  7.425e-06  -1.108  0.26797    
mgr          6.808e-04  5.401e-04   1.260  0.20749    
edu          2.646e+00  8.689e-01   3.045  0.00233 ** 
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for binomial family taken to be 1)

    Null deviance: 398.17  on 355  degrees of freedom
Residual deviance: 381.21  on 352  degrees of freedom
  (4 observations deleted due to missingness)
AIC: 389.21

Number of Fisher Scoring iterations: 4

and the model that is being graphed here is:

logit_edu<-glm(status~edu, data=Rust_Data_10_13, family= binomial)

I have also graphed a zoomed in version of the model but you cannot see the ends of the curve.

Can anyone help explain why this is happening/is my model just bad?

enter image description here

New contributor
vidada is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
  • 1
    $\begingroup$ "I have also graphed a zoomed in version of the model but you cannot see the ends of the curve." -- There are no ends of the curve! No matter how large a value $x$ you choose, you can compute $\sigma(x)$, and it will be between 0 and 1. $\endgroup$ – Sycorax May 2 at 23:44
  • $\begingroup$ Your graph is labelled MHI rather than mgr but no matter. You have what look like small coefficients in your multivariate logistic regression and these are described as not significant, so you should not expect these two variables to have much effect on your log-odds and your (univariate) graph suggests they do not $\endgroup$ – Henry May 3 at 0:53
  • $\begingroup$ Changed the post to have the EDU graph- still is clustered in the center, any suggestions? $\endgroup$ – vidada May 3 at 1:22
  • $\begingroup$ Your plot is nearly meaningless because it examines the relationship of the outcome with only one of the three explanatory variables in the model. Given the other variables are not significant, consider refitting the model with only edu as the explanatory variable. Then it would be meaningful to make such a plot: but it's pointless to extend the horizontal axis beyond the range of the data; that only makes the plot impossible to interpret. $\endgroup$ – whuber May 3 at 12:26

Your Answer

vidada is a new contributor. Be nice, and check out our Code of Conduct.

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

Browse other questions tagged or ask your own question.