Estimating Mean and Variance with Measurement Limits
Author Information
Author(s): Markéta Janošová, Stanislav Katina, Jozef Hanes
Primary Institution: Masaryk University
Hypothesis
How do different estimation methods affect the accuracy of mean and variance when observations fall outside measurable ranges?
Conclusion
Using a normal distribution with censored observations is recommended for estimating mean and standard deviation when less than 40% of observations are outside measurable ranges.
Supporting Evidence
- The study recommends using a normal distribution with censored observations if less than 40% of data is outside measurable ranges.
- Methods that ignore or improperly handle truncated data can lead to significant biases in estimates.
- Real-life data applications showed that different methods yield varying results based on the proportion of truncated observations.
Takeaway
When scientists measure things, sometimes the numbers are too high or too low to see. This study shows the best ways to guess the average and spread of those numbers when that happens.
Methodology
The study used simulation methods to compare four different estimation techniques for handling data outside measurable ranges.
Potential Biases
Using incorrect methods can lead to biased estimates and affect research outcomes.
Limitations
The study advises against using data with more than 40% of observations outside measurable ranges.
Participant Demographics
Data included various laboratory measurements from clinical trials.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website