DQE+Review+Questions+for+Van+Dorp

DQE Review Questions for Van Dorp (DQE Prep) (GWU EMSE-271)
Index | Topics (Logical Lectures) | Lectures | Problems | Readings | Nomenclature | Concepts

Top Concepts | Fill in the blank calculations | Academic Correctness | Notes | Other Oohs and Aahs | Test Question Bank | Problem-Solving Oriented Checksheets | Q for VD | Definititions (TBD) | Symbols (TBD) | Potential Question Analysis (Mulitvariate Analysis) | Classical Statistical Inference | 2011 Study Plan | Problems Index ]

Curent (January 2011) Questions for Van Dorp (Philip and Phillip)


 * [ Don't ask . Nice to know but not timeline critical.] What is the difference between Model Adequacy and Validation? For Regression, PCA and ANOVA?
 * [Don't ask, timeline priorities .] Should a Standard Error of Forecast be computed in all regression forecasting?
 * [Don't ask, timeline priorities .] Do Studentized and DFITS influential observations (residual analysis) tests only apply to normalized data? We think so.

List for January 2010 Team Meeting with Van Dorp

Christyl has the lead on the controversial questions.
 * [**Do we need to ask this part of the question?]** Would you contrast and compare Condidence versus Credibility intervalsagain.
 * [**Don't ask, we know the answer**.] Slides 103 1nd 104 (EMSE 271, Spring 2010) show both Confidence Intervals and Credibility Intervals (100 and 500 samples). Slide 104 states the width of the credibility has not changed. In experimenting with tab in the excel spreadsheet that demostrated the effects of a sample size, the Credibility Interval moved around, but it was not clear that the size did not change. From the lecture slides, should we assume that dispite how the credibility interval is displayed, the width remains the same?
 * Must be, a credibility interval lower bound is computed a the nth percentile of the distribution and teh upper bound is compute at teh 1-nth percentile of the distribuiton, the values won't change with sample size.
 * Will we need to calculate the credibiliy interval? Is there a way to do it without taking the inverse of the desired distribution?
 * **[OK]** In regression, when computing n-p-1 for the regression, do you include the intercept in determining the number of variables (p)? If not, why not? n-(p+1) p=number of independent variables + intercept.
 * In regression, when does an adjustment for sampling and modeling error need to be performed?When the values being estimated are farther from the mean? When the independent variables being estimated are outside the range of the data sample? Always? [Standard Error of Forecast]
 * [**Dont' ask, we know answer.]** When the dependent variable has been normalized I.e. Leslie Salt)? Nope the last adjustment was for transformation issues, see Lattin, page 70.
 * Why did we normalize the dependent variable in Leslie Salt? What could we have concluded if we didn't?
 * What is the solution for question 3 on the final exam? (2K ANOVA)
 * In ANOVA, are the definitions, computation and model validation the same in 1-Way ANOVA as in other ANOVA methods 2-Way and 2K Factorial? What are the similarities and differences between 1-Way ANOVA contrasts and 2K Factorial? (2X2)?
 * Why didn't we, in the course) do a Bartletts's Equality of Variance Test for regression and 2K Factorial ANOVA as was done for 1-Way ANOVA, 2-Way ANOVA, and PCA? Test performed were easy (easier than the tests not done) and sufficient?
 * What is the correct way to state (Academic Correctness) topics on the DQE?
 * Ask Van Dorp about the topics on the DQE test.
 * **[Don't ask]** Do we need to be able to compute simple matrix operations?
 * How long is the test? (2 hours)
 * Do we need the ln function on the calcuator?
 * How many questions on the test?
 * ?VD: Why is the null hypothesis = normal distribution, which is something we want? don’t we usually want to reject the null hypothesis to achieve a desired outcome?
 * [**Should We Ask**] Do we need to know Method-of-Moments | Maximum Likelihood Estimation?


 * Sources:**
 * EMSE 271, Fall 2009