How to interpret Weibull Accelerated Failure Time (AFT)

#1
Hello,

I fitted a Weibull Accelerated Failure Time (AFT) to my dataset (n=1071). The output reported the shape parameter (B=0.67,p<0.001). My understanding is that since the coefficient is less than 1, this means that the hazard is increasing over time. Is this correct?
I also fitted another Weibull AFT to a subset of the same dataset (n=584). The output reported the shape parameter (B=0.70, No p-value). Why isn't there a p-value? can I interpret this shape parameter without a p-value?

Thank you
 

Miner

TS Contributor
#2
No, a shape parameter (beta) < 1 is a decreasing failure rate (infant mortality). Beta = 1 is constant failure rate (useful life). Beta > 1 is an increasing failure rate (wear out). We cannot answer the second question without more detail. on how the analysis was structured. Simply sub-setting the data should not have caused that problem.
 
#4
No, a shape parameter (beta) < 1 is a decreasing failure rate (infant mortality). Beta = 1 is constant failure rate (useful life). Beta > 1 is an increasing failure rate (wear out). We cannot answer the second question without more detail. on how the analysis was structured. Simply sub-setting the data should not have caused that problem.
Thank you for your answer for the first question.
For the second question, my first model contained the entire population. The model consisted of the dependent variable (time to event), 5 categorical variables and 3 control variables. I ran the first model using the entire dataset with all variables included. Then, to conduct exploratory analysis, I filtered the original data using one level of my categorical variable. So my second model consisted of a subset of the original data (i.e., dependent variable (time to event), 4 categorical variables and 3 control variables).
 

Miner

TS Contributor
#12
Usually, when Minitab shows an * instead of a p-value, it means that your model is saturated (i.e., there are insufficient degrees of freedom to estimate the p-value). Look at your error term. If df = 0, that is the reason.