Statistics and data analysis for physics
In physics you often meet statistics through the lab course: a series of measurements, a fit, and an uncertainty that has to go in your report. I help you understand where those formulas come from, so you can also judge whether your fit is right.
What we work through
Measurement uncertainty
- Random and systematic errors
- Standard error of the mean
- Significant figures and notation
Error propagation
- Through partial derivatives
- Independent and correlated uncertainties
- Relative uncertainty for products and powers
Fitting
- Least squares and weighted fits
- χ² and reduced χ²
- Residuals and uncertainty in the fit parameters
Distributions and code
- Gauss, Poisson and binomial
- Counting experiments and √N
- Python: numpy, scipy curve_fit, matplotlib
Which courses?
- UvA/VU BSc Physics and Astronomy, e.g. a data analysis course or the first-year lab
- UvA BSc Chemistry or Science, e.g. a lab course with error analysis
- Applied Physics at a university of applied sciences, e.g. a measurement or lab course
These are examples (e.g.). Course names and content change from year to year, so check your own study guide. Send me your course and I will look at what you need.
Where it often goes wrong
Independent uncertainties add in quadrature
For z = x + y you get σ_z = √(σ_x² + σ_y²), not σ_x + σ_y. Adding them overestimates your error.
σ is not the uncertainty in the mean
Over N measurements that is σ/√N. More measurements make your mean more precise, not the single measurement.
A reduced χ² far below 1 is not good news
Around 1 is expected. Much smaller usually points to overestimated error bars, much larger to a poor model or underestimated errors.
What does it cost?
- I teach this subject myself: €300 online or €400 on location per 4-hour package, 1 student.
Two or more students can pay less per person, see the price table on the contact page.
Want to see if I can help?
The trial lesson is free. Send me a message with your course and when your exam is. I reply myself.