Argonauts:Introduction to Clinical Stats
From Wasteland
Contents |
Sun 30 April 2006
Spatio-temporal coordinates
WMVL, 11am to 1pm (exactly).
Attendees (in alphabetical order)
- Rehan (Bobby) Ali.
- Ramón Casero Cañas.
- Niranjan Joshi.
- Matt Kelly.
- Catherine White.
Minutes
O Spring! Time for Love and Clinical Statistics.
Hence the inauguration of the Second Journey of Discovery of the Argonauts (R-go-nuts). Our focus this Trinity term will be on
- Learning some Statistics that are relevant for Medical Image Processing.
- Learning the language of clinical publications.
- Working with an appropriate environment.
so that we can deal with questions such as:
- Is my algorithm better than his?
- What are the parameters that clinicians use in medical publications?
- What's the probability that this patient is ill?
- Does my algorithm work? Can it actually detect tumours?
- How can I model abnormal patients as a function of some variables?
- Are my results significant? Does my method make a difference?
- My algorithm has an error of 1mm. Is this good or bad?
This meeting was all about chatting about what we know already (that didn't take long) and what we can do this term. Some topics that were proposed:
- Probability Density Functions.
- Sensitivity, specificity, positive/negative predictive value, ROC...
- General Linear Models (special case ANOVA).
- Estimates, standard error, confidence intervals.
- Inter- and intraobserver agreement.
- Null-hypothesis testing.
- Outliers.
- Bootstrap statistics.
Intead of being merely theoretical sessions, the emphasis will be on practical training with real data. We are going to use the environment GNU R.
R is a language and environment for statistical computing and graphics. For us, it's like Matlab but specific for doing Statistics. R works under GNU/Linux and Windows, so that covers all needs in the lab. Besides, it's free software (as in freedom) and it's easy to find help online or from within the environment.
We installed the environment in 3 computers and started playing with it. The homework is to read the manual An Introduction to R. There is a pdf version as well as many other manuals in The R Manuals webpage.
--
