PHD Discussions Logo

Ask, Learn and Accelerate in your PhD Research

Question Icon Post Your Answer

Question Icon

What are the minimum number of measurements required to generate a statistically rigorous contoured stereographic diagram?

I've collected 25 fault plane measurements from a single outcrop to analyze the stress regime. My software will contour any dataset, but I know there's a point where the contour pattern becomes more noise than signal. Is there a rule of thumb or a statistical test (like a bootstrap) to determine if my N is sufficient to confidently interpret the clustered maxima, or should I keep measuring until the pattern stabilizes?

 

All Answers (1 Answers In All)

By Preetham M Answered 1 year ago

The old rule of thumb of N>30 is a decent start, but it's not about a fixed number it's about pattern stability. I use a pragmatic, two-step approach. First, I aim for a minimum of 40-50 independent measurements (avoid measuring the same fracture 10 times). Second, and more importantly, I perform a jackknife or bootstrap resampling test. I randomly subset my data (e.g., 80%) and re-contour it multiple times. If the primary cluster maxima jump around wildly, I need more data. If they remain stable in position and relative density, my N is sufficient. I've published with N=35 when the clustering was extremely tight, and rejected analyses with N=70 that were too scattered.

 

Your Answer