Advanced Stats exam

Absolutely no collaboration or communication of any kind with other students or friends to aid in the completion of these assignments is allowed. This includes, but is not limited to, verbal or email communication, text messaging, etc. If for any reason I deem that you have violated this policy, then you will at minimum receive a 0 for the assignment and most likely a failing grade in the course.

Please note: It is acceptable to email Dr. Nolan for clarifications related to the CUA – but please don’t expect answers to questions of a statistical nature about specific parts of this assignment. It is also important to remember that you may and should feel free to ask of me any general questions or questions about other assignments/feedback from the course and I will gladly answer those.

Please please please do not do anything that would cause me to need to enforce this policy!  Thanks!

Grading Rubric : Answers will receive…

Full Credit: Question is answered completely and correctly

80-95% Credit: Answer contains 1-2 minor statistical errors

60-80% Credit: Answer contains 3+ minor statistical errors or any major statistical error

0-60% Credit: Answer is substantially incomplete, contains multiple major statistical errors, or does not address the question that was asked.

Reminder : Expected answer lengths are given for many questions and reflect roughly the length of answers in my solution key. Please note that the expected length of answer will have nothing to do with the grading – however it is also true that the inclusion of irrelevant information (especially if it is incorrect) may result in deductions.

Questions 1-6 refer to Article 1, entitled Preterm Infant Thermal Responses to Caregiving Differ by Incubator Control Mode, found in the final cumulative assessment folder on Blackboard.

#1 (25 points) In Table 2, they provide means and standard deviations for birth weight for the two samples (air/skin).  First explain why the two samples are independent. Then use StatCrunch to conduct an appropriate test to determine if the mean birth weights differ between the two groups. Indicate validity conditions for the test and assess whether they are satisfied. Then give the appropriate output from StatCrunch and a sentence describing the results of the test. Also produce and interpret a confidence interval in an appropriate manner. Finally, indicate how this all might impact the authors analysis. (Expected length of answer:  SC Output + 150-250 words)

Hint: To conduct this test you should provide StatCrunch with “summary statistics”. You should not need to put anything into the data table.

#2 (12 points) In Table 5 a repeated measures, two-way ANOVA is conducted.  Explain WHY repeated measures is appropriate in this case.  Then identify the four factors involved in the study and their levels. (Expected length of answer:  100-200 words)

#3 (12 points) The means given in Table 4 can be placed into the interaction plot shown below.  Assuming that the larger differences in means that are observed in this plot are in fact significant, carefully interpret the information contained in the plot.  In particular, what are the most important effects?  The interaction effect tested significant in Table 5.  What does the plot tell you about its importance?  (Expected length of answer: 100-200 words) Note:  The vertical axis is temperature on the Celsius scale.  **Uses Article 2 I-plot.jpg**

#4 (12 points) In the discussion on page 643, the authors state: “Using t-test comparisons, abdominal temperature at care onset did not differ between control modes, although the mean air temperature in skin servo-control was slightly higher than in air servo-control (36.44 vs 36.37°C).” What three serious statistical mistakes have the authors committed in making this conclusion? (Expected Length of Answer: 50-100 words)

#5 (10 points) Based on the discussion section, identify two specific limitations of this study. Describe how a future study might avoid these limitations. (Expected Length of Answer: 100-200 words)

#6 (10 points) In a follow-up study, you wish to determine whether Apgar scores can predict the need for incubation. Write a brief introduction to a report in which you will analyze these. Make sure to include the statistical method you will use and any appropriate references.

Questions 7-11 refer to Article 2, entitled Association between heart rate variability and manual pulse rate, found in the final cumulative assessment folder on Blackboard.

#7 (10 points) The authors state in the caption of Table 2 (page 248) that they’ve used Pearson correlation for the first three items and Spearman correlation for Age and BMI. Based on all of the information they provide, do you agree with these choices? What additional items (that were not provided) should be reviewed to support these decisions? (Expected Length of Answer: 50-150 words)

#8 (10 points) Table 2 in a somewhat unusual manner provides 95% CI’s for the correlation. Consider the first line relating to the correlation between SDNN and Mean HR in HRV. Interpret the CI (or if you prefer, the squares). (Expected Length of Answer: 50-100 words)

#9 (10 points) On page 248 alone, the authors have reported no less than twenty p-values and CI’s. How should this fact affect their analysis? (Expected Length of Answer: <50 words)

#10 (12 points) In their analysis, the authors examined pairwise correlations and used simple linear regression. They likely should be using multiple regression. Explain how multiple regression might be used here and briefly discuss additional issues that might result from its use. Also indicate how such issues may have already affected their use of correlation and SLR. (Expected Length of Answer 50-100 words).

#11 (10 points) Near the end of page 248, the authors produced a 95% CI for the slope related to Pulse1 (0.80 to 1.00) in a model to estimate Mean HR in HRV as the dependent variable. Give the statistical interpretation of this interval. (Expected Length of Answer: <50 words)

Questions 12-15 refer to the StatCrunch dataset: 1601 Blood Gas Values

The goal based on these 150 observations is to predict arterial blood gas values (ABG, mm Hg) based on central venous blood gas values (VBG, mm Hg). You should use StatCrunch in coming up with your answers to these questions.

#12 (10 points) Assume that a regression is appropriate. Use StatCrunch to provide the estimate of R2 for the regression relationship and explain what the value means. (Expected length of answer: <25 words)

#13 (10 points) Assume that regression is appropriate. Obtain and interpret a 95% confidence interval for the slope. (Expected length of answer: <50 words words)

#14 (10 points) Assume that regression is appropriate. Use StatCrunch to predict with 90% confidence the ABG for a patient having VBG = 40 mg Hg. Provide a proper statistical statement from your prediction and then assess its clinical value. (Expected length of answer: 50-75 words)

#15 (10 points) Assess all assumptions of the regression model as best you can based on the data provided. If you use a graphic for your assessment, do not attempt to copy that graphic into Blackboard, but rather simply refer to it. (Expected length of answer: 50-75 words)

Questions 16-21 are unrelated to any particular article or dataset .

#16 (8 points) Elisabeth wishes to compare average heights across three different groups of people. She plans to collect data for 30 people chosen from each group. Provide Elisabeth with a brief indication of what procedures she ought to use in her analysis. (expected length of answer:  <25 words)

#17 (9 points) Explain why the following statement is false: When we lack evidence against our null hypothesis, this in general means that we should collect a larger sample in order to find such evidence (expected length of answer: 25-50 words).

#18 (9 points) The primary purpose of a manuscript I am reading is to present lots and lots of different correlations. Will I be able to make clinical use of the results? Why or why not? (expected length of answer: 25-50 words)

#19 (9 points) Which are more valuable in assessing the clinical importance of results: hypothesis testing or confidence intervals? Explain your answer. (expected length of answer: 25-50 words)

#20 (9 points) An author uses two-factor ANOVA to examine height as the response. He finds that there is an interaction between his two factors, gender and ethnicity. Explain, in basic terms, what it means to have such an interaction. You may use examples as appropriate (expected length of answer: 25-50 words).