-Subjects are characterized by cat. and con. var. -Subjects engage in a varying number of consecutive events -The number and kind of events subjects engage in depends to some extent on the subjects’ characteristics -Each event is characterized by cat. and con. var. -The dependent variable is "success", which is binary coded (1 = success / 0 = failure) -Series of events for each subject end either with success or failure. -I assume that the characteristics of an event depend to some extent on the characteristics of previous events and the subjects' characteristics. -I assume that the event series outcome (1 / 0) is influenced by the subjects‘ characteristics, the single events’ characteristics and characteristics of the event series.
I am trying to develop a model that predicts the probability of success for a new subject at any given event in the"event series". The dependent var. is measured at group level (individual subject = group of events).
-The group-level var. act as predictors for group-level outcome and the specific series of events (what are the events' characteristics, how are the events ordered, how many events are there) -The within-group variables (the event characteristics) act as predictors on subsequent event characteristics and the group-level outcome.
You have binary outcomes that influence each other, so that sounds like Markov modeling might be a suitable approach. You may want to look at UG ex 8.12 which can be extended to include covariates. Also study that literature. Perhaps also try SEMNET and/or Multilevelnet.