Test questions for next week

1. Please provide a brief summary describing how to critically interpret/read your advanced statistical topic when it is used in a scientific paper. Feel free to include a table or figure illustrating the guidelines, assumptions, or approaches to interpreting the reporting of this test in a scientific paper.

I really like this short article as an example: http://www.bmj.com/content/315/7109/672

2. Please repeat the above process but for a second topic that you did not present.

marking key /50 (but worth 25%) – please use as your outline
Q1. Worth 25 points.
Introduction to statistical test/10
Basic intro to purpose of statistical test & why you use it.
List assumptions and scope of inference for this test.
List advantages/disadvantages to this test relative to other options

Interpretation (or guidelines for readers)/10
Explain how to interpret the test-statistic(s)
Explain how to interpret the visualization
Explain what high, medium, and low values mean for the estimates and explain the accepted alpha for this test
List the key citations the reader would need to best understand this topic at a beginner level

Implications & context /5 (just a few sentences only)
Conclude your short note on this advanced statistical topic with a comment on its importance or changes over time. For instance, it is a relatively new technique and we can expect to see it more etc.
Implication or context could also include how it relates to our understanding of the underlying processes we are applying the statistical test to, i.e remind the reader the ‘why’ you do this test but in a bigger picture way here, for instance, this is an exploratory stat, or one for model building, or for assessing casualty, etc.

Q2. Worth 25 points.
Same marking key.

sample papers









What is it?

General Additive models, similar to GLMs and GLMMs but it mixes general linear models with additive models. Used for non-linear relationships.
They are good for maximizing the prediction of a dependant variable y from various distributions by estimating unspecific, non-parametric functions of the predictor variables.
How is it different from GLMs?
1. distribution of the dependant variable can be non-normal.
2. The dependant variable values are predicated from a linear combination of predictor variables.
Common Problems.
1. Results are not as easy to interpret as GLM and GLMMs
2. Very easy to “over fit” the data and not be able to replicate on future trials
3. Harder to summary and communicate to others
Heres a good website that explains GAMs well : http://www.statsoft.com/textbook/generalized-additive-models#additive%20models

unofficial blog for biostats