I regret I do not have adequate RAM (1GB) for use with S+ to gain any traction with this problem and that I did not discover the need to switch platforms soon enough (due to relocation). I did want to post a few basic questions:
Firstly, I notice that some PatientID's have only a single RecordID; my understanding is that a minimum of 2 were required and am suspicious that random partition into training and test sets was conducted at the RecordID, rather than PatientID level.
Secondly, I see an interest in predicting length of stay for specific outcomes from past visits (excluding the terminal visit altogether). What is your opinion of this?
I regret I do not have adequate RAM (1GB) for use with S+ to gain any traction with this problem and that I did not discover the need to switch platforms soon enough (due to relocation). I did want to post a few basic questions:
ReplyDeleteFirstly, I notice that some PatientID's have only a single RecordID; my understanding is that a minimum of 2 were required and am suspicious that random partition into training and test sets was conducted at the RecordID, rather than PatientID level.
Secondly, I see an interest in predicting length of stay for specific outcomes from past visits (excluding the terminal visit altogether). What is your opinion of this?