# LogLogistic Survival Model Assumptions

I am working with Hospital Length of stay data for the first time. It is highly right skewed. In researching ways to approach this problem, I thought a survival model fits the problem description.

The dependent variable is the waiting time until the occurrence of a well-defined event, observations are censored, in the sense that for some units the event of interest has not occurred at the time the data are analyzed, there are predictors or explanatory variables whose effect on the waiting time we wish to asses or control. http://data.princeton.edu/wws509/notes/c7.pdf

I was wondering what kind of assumptions go in to this kind of model, as opposed to say a Poisson regression, another alternative I was looking in to?

Is there any assumptions that need to be made about the distributions of the independent variables? What about how those variables change with respect to time?

Log-logistic expressed as hump-shaped hazard.

It shows that hazard has hump-curve at the initial timeline.

Hazard to be used:-

1) accident hump at age 18-21.

2) bypass surgery hump.(the risk of dying is more initially but after successful operation hazard decreases.)

Assumption:-

1) risk of dying is more at the initial timeline.(hazard increases)

2) after surviving the initial timeline risk of dying decreases.(hazard decreases)

Is there any assumptions that need to be made about the distributions of the independent variables?

Here, random variable is hazard at time and we took assumptions of hazards, now you need to find the parameters from large homogeneous samples to reduce sampling error and test the goodness-of-fit.

What about how those variables change with respect to time?

first increase until success then decrease.

• Is there any assumptions that need to be made about the distributions of the independent variables? What about how those variables change with respect to time? – user1775655 Apr 29 '15 at 19:57
• @user1775655 edited! – Hemant Rupani Apr 30 '15 at 6:05