Hello all, I would like to ask for your assistance with setting up a model using GLIMMIX. My data involves births in a hospital. The variable of interest is whether the mother stayed in the hospital motel or not (1/0). My data contains around 40,000 observations, from which almost 30,000 did not stay in the motel and the rest did. Each line in the dataset represents a birth (with 1 baby or more). Some rows represents births of the same mother, therefore I have a variable called MotherID, and this is the reason why I need a random effect logistic regression here. The independent variable list if quite long, with approximately 25 possible variables for me to choose from - some continuous and some categorical. I am now filtering some out due to high correlations between them. Some variables are birth related and some are mother related, for example: Socioeconomical status of the mother (numeric), number of babies in the current birth (1 to 4), birth number in this hospital (first, second, ...), did the mother take a birth preparation course (yes/no), was the mother in emergency room prior the birth (yes/no), pregnancy week in birth time (numerical), number of days in emergency room (numerical), etc,.... For the sake of this discussion, let's call the categorical variables C1, C2, .... and the numeric ones X1, X2, ... I wish to set up a model using GLIMMIX, a random effect logistic regression. I am not sure how to write it ( will give my attemp below ). In addition, the sample size is fairly large. This means that small effects will give significanct P-Values. I need to ask SAS to give me some measure of effect size, so I can know if a significant P-Value is actually interesting or not. I also want to produce probabilities of going to the motel, based on the variables in the final model. Regarding a final model, should I enter variables manually, or is there an automatic way (stepwise, etc...)? I am not too keen to count on automatic ways... My initial code is: proc glimmix data = motel method = quad;
class C1 C2 C3;
model Motel = X1 X2 X3 C1 C2 C3;
random int / subject = MotherID;
run; Is this code correct? How do I add effect sizes, predicted probabilities, confidence intervals, etc? Thank you in advance !
... View more
