Statistics Help @ Talk Stats Forum - Applied Statistics
http://www.talkstats.com/
Statistics in finance, economics, engineering. Actuarial science. Econometrics. Operations research.enSat, 28 Feb 2015 04:48:48 GMTvBulletin60http://www.talkstats.com/images/misc/rss.pngStatistics Help @ Talk Stats Forum - Applied Statistics
http://www.talkstats.com/
Holt Winters assumptions (aka exponential smoothing).
http://www.talkstats.com/showthread.php/59809-Holt-Winters-assumptions-(aka-exponential-smoothing).?goto=newpost
Fri, 27 Feb 2015 23:15:27 GMTI have not found assumptions of the methods for this raised very often except for outliers. And I have seen that it is very robust to violations of...I have not found assumptions of the methods for this raised very often except for outliers. And I have seen that it is very robust to violations of assumptions. So I have never tried to run a diagnoses.

Does anyone know what the assumptions of this are, how you check if they are violated, and how to address violations?
]]>Applied Statisticsnoetsihttp://www.talkstats.com/showthread.php/59809-Holt-Winters-assumptions-(aka-exponential-smoothing).Applying Credible Intervals vs Confidence Intervals
http://www.talkstats.com/showthread.php/59790-Applying-Credible-Intervals-vs-Confidence-Intervals?goto=newpost
Thu, 26 Feb 2015 07:24:19 GMTHi all,
Struggling through a problem at work, and we are slogging through an internal debate. We are not statisticians and rapidly finding the...Hi all,

Struggling through a problem at work, and we are slogging through an internal debate. We are not statisticians and rapidly finding the limitations of our knowledge in the field! I've spent about a day and a half researching this issue and am here seeking advice help.

I'll post the mock scenario below with my thoughts!

Scenario

I'm a college professor lecturing a course with 2000 students. On the first day, all students answered a 6 question test, where the difficulty of the questions amounts to "2+2+?". A team of TAs graded all 2000 exams and reported that everyone, as required to continue with the course, scored 6/6.

I want to know how many of these exams must I manually recheck to be confident (preferably 95%) that the remaining sample are correct.

Thoughts and Problems

The initial work done on this led to a frequentist approach using 95%CIs and setting the SD to 12, giving the number of tests to be regraded at ~65. There was no consideration for the # of questions on the exam (and I'm not sure if there should/should not be). My other larger issue with this approach was that the SD was arbitrarily set to obtain a manageable rechecking sample.

I also struggled knowing that my desired outcome is binary: I want to recheck a number that leaves me very confident that the remaining tests have been correctly graded. I have been flirting with the idea that using a credible interval here is a better method.

Does anyone have any insights into how I would begin using this method, if it is in fact at all in the correct direction?

Appreciate any thoughts or critiques!

AB
]]>Applied Statisticsabgtrhttp://www.talkstats.com/showthread.php/59790-Applying-Credible-Intervals-vs-Confidence-IntervalsAggregate Loss Model (Actuarial Science)
http://www.talkstats.com/showthread.php/59785-Aggregate-Loss-Model-(Actuarial-Science)?goto=newpost
Wed, 25 Feb 2015 17:09:14 GMTQuestion:
Let S be the random sum of N independent frequencies {{M_i}} where N is negative binomial distributed with r = 5 and \beta= 3 and {M_i}...Question:

Let be the random sum of independent frequencies {} where is negative binomial distributed with and and is Poisson distributed with = 3.

Find

Attempt:

I know the formulas for these two distribution, but I am totally lost of how I should approach this. A step by step guide would be much appreciated! Thank you!
]]>Applied StatisticsAsanohttp://www.talkstats.com/showthread.php/59785-Aggregate-Loss-Model-(Actuarial-Science)