Stat 512 - Final Project Notes

Final Reports: Your final reports are due on or before Dec 2. Reports must be typed, with relevant computer output incorporated into the body of the paper. Turn in one report per group and include previously read project reports. Raw data should be turned in on a separate well-labeled floppy disk or emailed to me. You will turn in project evaluations next week as well. You may assume your audience will understand all statistical terminology. See the “checklist for final reports” at the end of your syllabus. You don’t have to follow the section suggestions exactly, but I encourage use of section headings, and please make sure all of the information requested is there. Especially don’t forget to give me lots and lots of detail on how you collected the data. Please also include a brief “statement of purpose” that can be posted on the web for future classes to review.

Final Report DO’s:

· Make sure all graphs are clearly labeled and formatted (even number the figures).

· Remember to interpret your graphs and results.

· Summarize and format key information, don’t merely copy and paste Minitab output.

· Use full and complete sentences. Don’t use excessive jargon.

· After the first draft is complete, put it aside for at least 24 hours, then read it through as though seeing it for the first time. Review and revise. Spell check!

· Use symbol font for Greek letters and use the equation editor, can also use subscripts. Use “courier” font to align Minitab output. You may also have to reduce font size to 10 for it to fit nicely on the page. Output should never run over the end of the line or the end of page.

· To import graphs from Minitab, save them as .jpg files and then import them into Word (see green handout).

· Keep in mind that a statistically insignificant result is not a failure (!) but may actually be more interesting than finding what you expected or may result from small sample sizes.

· Include all the details in the report even though the presentations will be brief.

Project Presentations: Each presentation will be at most five minutes (this is harder than it sounds – you should practice the timing before hand!). Any number of group members may be involved in the presentation, this is completely at the discretion of the group. Feel free to be creative. The presentation should be done with PowerPoint or “transparencies” (in the studio you can project a regular piece of paper but formatted for presentation). For PowerPoint presentations, you can email them to me ahead of time or bring them on a floppy or a thumb drive. Groups have the option of volunteering to present Tuesday, contact me.

Presentation Order:

Project Presentations: DO’S

· Attached is a modified presentation from a previous year. It serves as an example of what your presentation can look like. I’m not expecting you to follow this “template” exactly, but to be reminded of issues of organization, effective presentation, brevity (highlight the most interesting points), and proper justification of conclusions. This presentation is an outline, embellished during the talk to fill in details.

· Begin by motivating your study. Why did you do it? Why was it interesting to you? What were you interested in finding out? What were your goals? What background information does the audience need to know?

· It is good practice to early on give an overview for the entire presentation (tell the audience what you are going to tell them during the presentation) but you won’t have much time for that here.

· Give details about how the data were collected. This is often the most interesting part of the presentation. Consider providing us with a copy of survey questions, etc.

· Consider summarizing which variables you plan to focus on and whether they were measured as quantitative or qualitative variables (e.g., variable 1 = time for order to be filled (quantitative), variable 2 = walk in vs. drive-thru (qualitative)).

· Provide graphical summaries of your data. This is very effective as a presentation aid and will help you discuss your technical conditions (is the sample normal looking?) and conclusions (was there a big difference between the two groups?) later. Don’t just tell us there is a difference or an association but indicate in which direction.

· Summarize your hypotheses statements and test of significance results in English. Report test statistic value and p-values. Consider combining the graphical summary with the test results so that we can see the difference/relationship. Link the descriptive with the inferential. Anytime you make graphical and numerical summaries you should pause and comment on what information they give you before you proceed to tests or intervals.

· Make sure you discuss briefly whether the technical conditions were met and how that affects the validity of your conclusions (include more details in the written report). Finish with cautions/recommendations for future study.

· Speak loud and clearly. You may be required to use a microphone! Speak to the audience, not the me, not to computers. Eye contact is important.

· Most of all – be creative! Be enthusiastic and have fun!

Project Presentations: DONT'S

· Never just copy and paste Minitab output. You need to select the most relevant information and present that in a clear, concise manner.

· Don’t make the presentation aids so “showy” that they overwhelm the statistical content. Don’t overcrowd the slides. Hit the highlights. Guideline: roughly one minute per slide.

· Don’t use small font. Aim for at least 24 point font.

· Don’t be too repetitive. You should hit the highlights and be ready to answer questions if anyone wants to know more detail.

· Don’t run overtime. You will be cut-off after 5 minutes. You should rehearse ahead of time and practice the timing of the presentation. This will also reduce hesitations (um’s) during your presentation.

· Never read your presentation! You should also practice speaking slowly and clearly. Know what you want to say before you get up there.

· Don’t talk above or below your audience’s level of understanding. Know your audience.

· Avoid distractions, e.g., playing with coins in pockets.

· Don’t use words like “cause” and “affect” and “impact” with observational studies.

Audience responsibilities during presentations

· Pay attention. I often get good ideas for final exam questions from class presentations.

· Be respectful. Remember, you’ll all be up there too.

· Ask questions – this is your end of course review! Use these presentations as a valuable review and learning experience!

· Evaluation – the audience will evaluate the presentations on

o introduction: was a good overview of project provided? captured your attention?

o organization: was the presentation easy to follow? was it well organized?

o clarity: was the presentation clear and understandable?

o visual aids: did they enhance the talk? were they clear? were there enough?

o conclusion: did the group present the final results, recommendations, future questions, and wrap everything up clearly?

Comments that were made to students previously after their presentations:

· Don’t forget about including confidence intervals to estimate the size of the “effect.”

· Don’t say “number of samples” when you mean “number of observations in the sample.”

· If we fail to reject H₀, this means there is a lack of evidence against the null hypothesis. It is never correct to say we have evidence in favor of H₀. If we reject H₀, this means we have evidence against the null hypothesis (we are always making a decision about H₀).

· We never “prove” anything, we just have evidence (or don’t have evidence)…

· The p-value tells us how strong the evidence is. A p-value of .0001 is very strong evidence against H₀. A p-value < .05 is moderate evidence against H₀. A p-value of .06 doesn’t mean we have no evidence against H₀, just not quite enough to convince us to reject it (if using .05 as our standard).

· The form of the alternative (<, >, ¹) tells us what types of observations we will consider as evidence against H₀. If H_a uses <, then the p-value is the area to the left of our test statistic. If H_a uses >, then the p-value is the area to the right of our test statistic. Note, these are not always the smallest areas. If we guessed wrong in H_a, then we get a p-value > .5. If H_a uses ¹, then we look at both tail areas and our p-value is twice as large as a one-sided p-value. It is crucial that you use the correct direction in calculating your p-value and in using Minitab. A sketch of the sampling distribution is very helpful here. If you had a suspected direction (in H_a), you should use that one sided p-value, not the two-sided p-value.

· You seem to be throwing around the term “bias” a little too easily. There are always going to be other variables around. These are not sources of bias until you can establish a ‘systematic difference’ between the sample results and the “truth.” For example, in the marijuana study, the group thought that people with lower GPAs might be less likely to participate in the survey. This would make the sample mean GPA results they got higher than the population mean, even if they took a larger sample (it’s the reasons for not responding that matter, not the number of respondents). The roommate group commented that freshmen who are in their room a lot were more likely to end up in their sample than people who are seldom in their room. So when they asked “how much time do you spend in your room,” the results they get will have a tendency to be longer on average than the population as a whole. There are always sources of variation in the data, that’s why we have statistics, but that doesn’t automatically mean there is bias. You should be specific in the direction of the bias if you are suspecting bias.

· Similarly, don’t overuse the term “confounding.” There will always be other variables, but they are not confounding unless they work differently between your treatment groups (e.g., cold weather and holiday season – two things change at a time). Don’t use “confounding” to refer to sampling errors. Remember bias describes a potential problem when we want our sample to represent the population and confounding describes a potential problem with dissimilar treatment groups.

· Increasing the sample size does not decrease bias and it does not matter what proportion of the population you have taken in determining bias or precision. (Though may need to make some adjustments if population size is not huge – see me.)

· Remember to check whether your samples are independent before you do any of the two (or more) sample procedures.

Other cautions:

· affect vs. effect; there vs. their; data are plural; it’s vs. its