Survival data is tricky to model. They have just done an overall average number of failures per running hour. There do seem to be a lot of factors not taken into account here such as drive ages, (in both running time and elapsed), etc.
With right censored data (as this is), if you measure age at death but then you're only modelling already failed drives, so you'll under represent good drives.
It would be good to see some statistics done so we can see confidence intervals around a hazard rate at different ages.
With right censored data (as this is), if you measure age at death but then you're only modelling already failed drives, so you'll under represent good drives.
It would be good to see some statistics done so we can see confidence intervals around a hazard rate at different ages.