Good intro to KM survival stats. I appreciate the author highlighting the limitations of KM due to non-parametric results because of discrete/binned data, such as "number of machines that failed per month".
IMHO it's well worth also knowing about parametric results using continuous data, especially if you're involved in tech products, because tech product telemetry can often provide individual timestamps, such as "this specific phone worked at $timestamp and didn't work after".
It would be nice if Amazon integrated a failure report feature, so we could get structured data on how long things last and what failures happen at what time.
There's reviews, but unless you manually read every one they're not useful, most bad reviews are things like "the motor isn't very powerful" on something that doesn't need to be powerful, or "it feels cheap and poorly made" because someone prefers metal over plastic.
This is hard to do for products when they change over time. Eg if a product is upgraded and repaired, and if it has different versions and so on, itβs no longer a static distribution, and things get complicated.
IMHO it's well worth also knowing about parametric results using continuous data, especially if you're involved in tech products, because tech product telemetry can often provide individual timestamps, such as "this specific phone worked at $timestamp and didn't work after".
A good place to start for the continuous approach is with the Weibull distribution: https://en.wikipedia.org/wiki/Weibull_distribution
reply