[R] Interpretation of Coefficients of GAM
thatsmile333 at hotmail.com
Thu Mar 17 00:07:03 CET 2016
May I have a question on how to interpret the coefficients from GAM.
We used GAM to get propensity scores, fitting model as below,
fit <- gam(group~agegroup+s(GA_Enr)+race+edugroup,
data = datax, family='binomial')
where group is a binary variable for treatment(y/n)
agegroup, race, edugroup are categorical
variables for age, race, and education years respectively.
GA_Enr is a continuous variable for age at enrollment.
We added smoothing term to it.
And we got the propensity scores as below,
ps <- predict (fit, data = datax)
Next we tried to manually calculate Pscores using the
coefficients from GAM. Our purpose was to compare the ps we manually calculated with the
ps we obtained from the above line. They are supposed to be the same.
Both coef(fit) and fit$smooth show coefficients from GAM.
Now for GA_Enr, we have one general coefficient
from coef(fit) , here call it coeff1.
And still for GA_Enr, from fit$smooth, we have
coefficients for each subject, here call it coeff2.
My question is how to interpret coeff1 and coeff2 and how to
use them to calculate ps for each subject?
Thank you so much in advance!
[[alternative HTML version deleted]]
More information about the R-help