Keeping your machines operating smoothly in order to have quality products as well as having them delivered in the proper timeframe to keep your customers happy is a cornerstone of every successful manufacturer. It is unavoidable that even the best machines will eventually need replacing, so it is ideal to know the probability of when your machines are likely to fail. To get a good idea of what the life expectancy of your machines will be before needing to be replaced, you can use the Kaplan-Meier estimator.
Overview: What is Kaplan-Meier?
Kaplan-Meier is a part of survival analysis in statistics used in estimating the survival probability of a subject from the time of a significant event.
The Kaplan-Meier formula is represented as St+1 = St*((Nt+1-Dt+1)/Nt+1)
3 drawbacks of Kaplan-Meier
While there are clear benefits to the Kaplan-Meier method, such as the graphical presentation being intuitive, there are some drawbacks to consider:
1. Easily biased
With the Kaplan-Meier method being invariable, this can lead to easy biasing. Reportedly, nearly half of the analyses utilizing Kaplan-Meier may overestimate risk due to competing risk bias.
2. Limitation
With the Kaplan-Meier being able to understand only a single factor at a time, it is not effective for the analysis of multiple variants.
3. Some aspects are unpredictable
The Kaplan-Meier method of analysis is unable to gauge the proper estimation of the size of the event making the change.
Why is Kaplan-Meier important to understand?
The Kaplan-Meier estimator is important to understand as it can be very useful in business and manufacturing as an estimator for a number of different things. We already know that it can be used to gauge when machines might fail, but here are some other ways to use this estimator:
Estimate critical WIP level
The Kaplan-Meier estimator can be used with a line that has no variability when an estimate is desired for when the maximum throughput is achieved at the minimum cycling time.
Employee turnover
With the Kaplan-Meier estimator, you can estimate the time from an employee’s hiring to when they will likely quit or be terminated.
A salesperson’s success
You can estimate the time from a salesperson’s hiring to when their first sale is likely to be.
An industry example of Kaplan-Meier
A soap plant is deciding on when the electric machines used in manufacturing their soap are likely to break down. Using the Kaplan-Meier method, it is determined that of the three main electric machines used in their soap manufacturing, Machine A has a survival probability at 10,000 hours of 95%. This drops sharply by 25,000 hours, where the survival probability is 35%. Machine B has a survival probability of 87% at 10,000 hours and 65% at 25,000 hours. Machine C has a survival probability of 98% at 10,000 hours and 76% at 25,000 hours. With this information, the plant is able to plan for when they will likely need to invest in replacements for which machines.
3 best practices when thinking about Kaplan Meier
When utilizing Kaplan Meier, there are aspects that are important to take into account:
1. Testing the results for more than one sample
In order to make any proper conclusions from the estimates reached using the Kaplan-Meier method if you have more than one sample, the log-rank test should be utilized. This test establishes whether there is a differentiation between the survival distributions of two groups.
2. Be aware of the deletion of censored data
When you delete censored data, the result is a change in the curve’s shape.
3. Do not rely solely on the Kaplan-Meier method
The results from the Kaplan Meier method can be easily biased. It is a great tool, but room for other methods should be encouraged to get the full picture.
Frequently Asked Questions (FAQ) about Kaplan-Meier
What is the difference between Cox regression and Kaplan Meier?
Cox regression allows for multiple predictors and is semi-parametric. Kaplan-Meier only allows for one predictor and is non-parametric.
What are the assumptions of Kaplan-Meier?
There are three assumptions in Kaplan-Meier survival. The first assumption is that the survival probabilities are the same at whatever time of various samples’ introduction into the study. Secondly, it is assumed that the event occurs at the time specified. Lastly, it is assumed that the survival probability is the same for samples that have been censored.
What is the product-limit estimator?
It is a name for the Kaplan-Meier estimator.
Proper planning in your business
Being able to estimate when an event may occur is important for the planning of your business. Things happen, but with tools like the Kaplan-Meier estimator, you can be better prepared for when they do.
