Saturday, June 25, 2016
Interesting write up of the importance of using diagrams and plots (but not photos or equations) in achieving highly cited publications...
Posted by Connie at 6/25/2016 06:21:00 PM
Thursday, June 23, 2016
The draft of The GRIM test: A simple technique detects numerous anomalies in the reporting of results in psychology describes a simple way to check for errors in small studies that collect data using integer scales and report means.
It's received some attention in popular press (like this article in the Economist) and has been posted on PeerJ Preprint -- "What is a PeerJ Preprint? A PeerJ Preprint is a draft of an article, abstract, or poster that has not yet been peer-reviewed for formal publication. Submit a draft, incomplete, or final version of your work for free."
The draft includes links to their project on Open Science Framework -- where one can access a spreadsheet to run calculations and the data presented in the article.
The discussion on the post includes a link to plain-language explanation, a reference to "Student's" original work, and a web-implementation of the test (created by a reader to run the numbers and visualize the outcome).
It's an interesting example of the use of open science tools as well as an interesting way to check validity of reported results in small studies.
Posted by Kairn Stetler Kelley at 6/23/2016 03:37:00 PM
Wednesday, June 15, 2016
Present: Nancy Gell, Juvena Hitt, Kairn Kelley, Ben Littenberg, Gail Rose, Connie van Eeghen
Start Up: It may be allergy season – symptoms were shared by many.
1. Discussion: Ben on how to teach recursive partitioning via classification and regression trees
a. Regression is a form of categorization
i. We see this in branching schemes, such as KPCOFGS branching diagrams (e.g. species of monkeys as the terminal points of this branch
ii. Diagnostic testing, e.g. determining CKD by GFR < 60 over 3 months; if >= 60 determined by urine test ACR <= 30
1. 10 % of population with CKD
a. 6% positive GFR (60/1000)
b. 94% negative GFR
i. 40 positive for ACR
ii. 0 negative for ACR
b. Prediction model, for CKD in the next 5 years (binary outcome) (Note: this is not a decision tree – it is predictive, not normative)
i. GFR in the 60-90 range (close to being low)
1. Positive: 50/300 (17%) CKD
a. High blood pressure
i. Positive: 45/150 (30%)
ii. Negative: 5/150 (3%)
2. Negative: 50/700 (7%) CKD
a. Age < 50 (note: not high blood pressure; every interaction can be unique)
i. Positive: 5/500 (1%)
ii. Negative 45/200 (23%)
1. Sex (M/F)
a. M: 40/100 (40%)
b. F: 5/100 (5%)
ii. Note: every terminal branch is either greater than 25% or less than 5% (rule created for this example)
c. Enter a patient
i. GFR is between 60-90 and High blood pressure is high: 30% probability
i. How to pick variables? The bigger the difference the better variable.
ii. How to decide when to stop – what determines the rule? When the difference between two terminal points is the greatest.
iii. If a continuous variable, how do you bifurcate? Don’t have to; OK to have more than two outcomes.
e. Demonstration: build a prediction model using one set of cards to predict outcomes in another
i. This model of learning could also be used to understand specificity and sensitivity
a. June 17, 2016: Gail – manuscript on IVR
Posted by Connie at 6/15/2016 02:53:00 PM