Comparing ANOVA results for glm vs. gam models ?

Question

[pierre.jorcin]

Hi everyone,

I am struggling to compare models performance, to evaluate results from glm vs. gam models. Using the below formula, I am trying to evaluate ANOVA results for both models, but results indicators are different. How can I interpret them, individually first, then one vs. another ? best regards,

m <-mGAM@models$presence$gam$`1`@object
m
anova(m)

Family: binomial 
Link function: logit

Estimated degrees of freedom:
8.88 6.42 6.50 6.18 4.71 1.00 3.11 
2.46 7.54  total = 47.8


UBRE score: -0.3067535  

Family: binomial 
Link function: logit 

Formula:
presence ~ s(Chelsea_bio4_R) + s(Chelsea_bio9_R) + s(Chelsea_bio10_R) + 
    s(Chelsea_bio14_R) + s(Chelsea_bio15_R) + s(Chelsea_bio16_R) + 
    s(alt_R) + s(orient_R) + s(NDVI_july_R)

Approximate significance of smooth terms:
                     edf Ref.df  Chi.sq  p-value
s(Chelsea_bio4_R)  8.882  8.994 102.044  < 2e-16
s(Chelsea_bio9_R)  6.422  7.270  36.608 8.97e-06
s(Chelsea_bio10_R) 6.501  6.774  66.955 1.52e-09
s(Chelsea_bio14_R) 6.184  7.099  26.832  0.00045
s(Chelsea_bio15_R) 4.707  5.858   8.455  0.20454
s(Chelsea_bio16_R) 1.000  1.000   3.855  0.04961
s(alt)           3.106  3.925   3.890  0.32081
s(orient_R)        2.457  3.064   9.847  0.02181
s(NDVI_july_R)     7.541  8.469  22.067  0.00554
  

m <-mGLM@models$presence$glm$`1`@object
m
anova(m)

Coefficients:
    (Intercept)   Chelsea_bio4_R   Chelsea_bio9_R  Chelsea_bio10_R  Chelsea_bio14_R  Chelsea_bio15_R  Chelsea_bio16_R  
     -6.6081112        0.0037166        0.1217509        0.1836676       -0.0526189       -0.0024913        0.0064658  
          alt_R         orient_R      NDVI_july_R  
     -0.0004764        0.0023634       -0.0002167  

Degrees of Freedom: 2159 Total (i.e. Null);  2150 Residual
Null Deviance:     2848 
Residual Deviance: 1782        AIC: 1802

Analysis of Deviance Table

Model: binomial, link: logit

Response: presence

Terms added sequentially (first to last)


                Df Deviance Resid. Df Resid. Dev
NULL                             2159     2847.5
Chelsea_bio4_R   1     1.02      2158     2846.5
Chelsea_bio9_R   1   704.34      2157     2142.2
Chelsea_bio10_R  1   196.65      2156     1945.5
Chelsea_bio14_R  1   103.93      2155     1841.6
Chelsea_bio15_R  1    24.30      2154     1817.3
Chelsea_bio16_R  1     3.82      2153     1813.5
alt_R            1     0.57      2152     1812.9
orient_R         1    15.17      2151     1797.7
NDVI_july_R      1    15.26      2150     1782.5
  

[h.mon]

I have tried to format your post to make it more readable, I hope I didn't delete information in the process.

Your question is more suited to Cross Validated. Be sure to do some reading before posting there:

How to compare GLM and GAM models

GAM versus GLM: same fit, different significance of predictors

When to use a GAM vs GLM

ANOVA to compare models

comparing models in R