Finding statistically significant "Outliers" from sequential data I have a need to find data points whose values are statistically different (significant) from sequential data points. For example I'm looking at weekly data points and as new data points are added I want to know if the point(s) added are statistically significant different from the group.  
Bernard Rosner (May 1983), Percentage Points for a Generalized ESD Many-Outlier Procedure, Technometrics, 25(2), pp. 165-172 has what seems to be a promising method.  
Any other suggestions? 
 A: Sequential models, changepoint detection and analysis is an entire class of methods, kind of unto themselves. Developed by Abraham Wald who published a wonderful, intuitive book titled Sequential Analysis in 1947, it's kind of languished in Op Res departments ever since. It has only been in recent years with Google's adoption of SPRT tests (sequential probability ratio tests) that it has emerged or come into its own. 
Recent publications include last year's Sequential Analysis by Alexander Tartakovsky -- a comprehensive look at the field.
http://www.amazon.com/Sequential-Analysis-Hypothesis-Changepoint-Probability/dp/1439838208/ref=sr_1_3?ie=UTF8&qid=1445907717&sr=8-3&keywords=wald+sequential+analysis
In addition, last June Columbia co-sponsored a 3-day workshop about it... https://sites.google.com/site/iwsm2015/committees-and-sponsers 
The OP's query is concerned with obtaining "suggestions" for what is clearly a question about changepoint detection in a time series. It deserves a more specific answer than "look within these references as there is sure to be a solution for your concern," but that will have to do for now. I'm interested in what others have to say.
