Statistics 309H

FAQ: Group Projects

How large are the groups?

There are 9 groups in each of the three classes (about 40-45 students per class). This means that there will be 4-5 people per group.

Can we form our own groups?

If you have a strong preference for working on your group project with one or two others in the class, send Jennifer (the project TA) an e-mail by 6:00 on Wednesday, Jan. 25th. If you do make a request, Jennifer must have an e-mail from all parties indicating this preference. We will try to accommodate these requests but we make no guarantees.

What should we do if we can't find or agree upon a topic?

Identifying a worthwhile, interesting, and potentially educational topic is part of your job in this class. Please don't ask Jennifer or me to give your group a topic or to decide among several choices.

Can we study topics that we discussed in class?

You can, but experience indicates that these projects usually do not score particularly well on originality. Spend some time up front thinking of interesting, original topics.

Can you provide us with examples of group projects from the past?

Sure. Here are a bunch, framed in question form. (Obviously you should not repeat any of these.)

• Can experts predict the stock market?
• Does cell phone usage cause accidents?
• Does personality type predict preference for [you fill in the blank]?
• Does the season in which a relationship started predict its longevity?
• Does the E-bus encourage student drinking?
• Is racial profiling used by airport security?
• Can anyone predict the success of Broadway musicals?
• Does confidence affect test performance?
• Why do some like Leno while others like Letterman?
• Are right-brained people more creative than left-brainers?
• Should fast food customers drive-thru or walk-in?
• What is the density and variability of raisins in Raisin Bran cereals?
• What factors influence attractiveness?
• When does background music improve performance?
• What is the optimal strategy in rock, papers, scissors?
• Does the hot hand exist?
• Do winning sports teams improve the city's economy?
• Who will receive the death penalty?
• What variables predict bathroom stall choice and hand washing behavior?
• Do good liars make good leaders?
• Can we trust polygraphs?
• Are symmetrical faces more attractive?
• Who will touch wet paint?
• Are some airlines really safer than others?
• Do prophets exist?
• What is the probability of transmitting HIV in the boxing ring?
• Will I become a parent before my time?
• Does anyone have ESP?
• Are good researchers also good teachers?
• Is there a price/quality relationship in the retail goods market?
• What are the safety implications of changes in the speed limit?
• Are women in accounting firms underpaid?
• How safe is safe sex?
• Does shelf placement affect sales?
• Do UT coaches understand "regression to the mean" in athletic performance?
• Is 98.6 a normal body temperature?
• How well do people play blackjack?
• Are the point spreads in professional sports biased in favor of big city teams?
• Do times and dates of death occur at random?
• How likely is it that an athlete who tests positive for cocaine is actually a cocaine user?
• Do blondes have more fun?
• Do baseball players who earn more perform better?
• Is sacrifice bunting a good strategy in baseball?
• Can people predict basketball players' shots?
• Which gum lasts the longest?
• Are men who smell good more attractive?
• Does caffeine keep you up or is it just a placebo effect?
• When will people pick up spare change?
• Does bottled water taste better than tap water?
• What factors predict winning an Academy Award?
• Do people name their babies after TV characters?
• Do crimes by minorities get disproportionate attention in newspapers?

Do you think that THIS topic will make for a good presentation?

Most topics are potentially O.K. so long as they lend themselves to some of the concepts and methods discussed in class. The first trick to having a strong presentation is to have a solid underlying theory that directs the research hypotheses.

For example, let's say that you wish to test whether hairy men are more likely to stare at themselves in the mirror than less hairy men. Before running this experiment, you should identify one or more theories that make this a theoretically compelling study. That is, why would "hairiness" tend to predict mirror staring?

The second trick is to have "interesting" hypotheses. Is your hypothesis so obviously true or false that it doesn't really pay to investigate it? Can you identify the conditions under which it is more and less likely to hold?

What do you mean when you say "have 'interesting' hypotheses"?

A study is "interesting" if, after hearing about the theory and results, audience members walk away saying, "Wow, I can't wait to tell some of my smartest friends about this one!"

Can you provide some specific guidance for identifying interesting hypotheses?

First, you should spend some time thinking about topics what would make for a fabulous presentation.  This is arguably the most important stage in your project.  If you pick a lackluster topic or method, your classmates might not think highly of your project no matter how hard you work on your study.  Some curiosity, creativity, and ingenuity is crucial. 

If you're interested in fashion, you might think about measuring the hemlines of dresses manufactured in different years to test your theory that hemlines come up when the economy is strong and down when it is weak.  If you're interested in sports, you might use Z scores to figure out who was the greatest athlete of his/her day.  If you're interested in animals, you might run a study at town lake to see whether dog owners look like their dogs. 

When predicting differences between groups, a "conditions under which" hypothesis is usually most interesting from a scientific standpoint.  See Jennifer for more information about this.

Are there any topics we should avoid?

Yes. Avoid all 'sensitive' topics. This includes grades, test scores, starting salaries, and anything else people are likely to lie about when they're on the unflattering end of the spectrum.

Also, avoid anything that compares students across majors and colleges. These types of comparisons are too often trite, unoriginal, and obvious.

A good example of a study that should be avoided is one that proposes that business students are more interested in money than theatre students. Of course they are! Likewise, it is true that students in psychology are more interested in talking out problems with their friends than students in engineering. Studies that pursue this type of issue are really just showing that there are differences precisely where everyone would expect them.

Also, avoid any topic that requires the collection of data that you are unlikely to be able to secure in time. For example, do not propose to compare the proportion of spoiled election ballots throughout Texas as a function of political orientation unless you know that you can actually get the ballots (unlikely).

What about looking at differences between men and women?

In general, studies of gender differences should be avoided for similar reasons to those given above. Men and women are different across a range of variables and everyone knows this.  Identifying what these differences are is not particularly interesting from a scientific standpoint.

A more interesting scientific question is "What drives a particular gender difference?" You're on the right track when you're able to say: "THIS type of man/woman (or men/women in THIS condition) will be Xs, but THIS OTHER type of man/woman (or men/women in THIS OTHER condition) will be Ys."

For example, if you think men are more aggressive than women, your theory might be that this is true because men learn their aggression from the male role models (i.e., their fathers). If so, then men who don't have male role models should not be particularly aggressive. So you might run the following test: compare the aggression of (a) men who have male role models, (b) men who don't have male role models, and (c) women. (In reality it may be hard to separate men into these groups, but this is just one idea.) You might predict that A > B=C. Do you see how this is a more "interesting" study than one that merely predicts a gender difference? It accounts both for the difference (most men are probably in group A), and it provides evidence about why it might be so.

Will you meet with our group?

Yes. I will meet with all groups in April (see your Syllabus). But you should plan this meeting carefully to get the most out of it. Before meeting with me, you should (a) spend a lot of time meeting as a group to explore possible topics and approaches, (b) make sure that every group member has thoroughly read this Group FAQ page, and (c) meet with Jennifer.  I have written detailed answers here so that we don't have to spend time going over these things during our meeting time.

There are 27 groups across the three stats classes.  This means that there won't be enough time for me to have multiple meetings with each group. Please prepare carefully for your group meeting with me.  Also, don't hand me a three page survey and say "This is what we're doing."  Instead, summarize your project for me before we get into specifics.  This summary should begin with a theoretical rationale for what you are doing, in addition to a description of your hypotheses.  If you don't know what I mean by "theoretical rationale," then you should reread the FAQs above.  

I have a quick question about our group project that I'd like to ask you right before or right after class.

There is no such thing as a quick group project question. The reason is that I can't remember what each of the 27 groups are doing at any given moment. When talking to you in my office, I'm completely focused on your project. But I am not focused on you when I'm getting ready for class or shuttling between classes. Please don't convince yourself that your question is so short and so simple that it's worth ignoring this request from me.  If you need to get feedback from me on a small matter, put your question in writing and send an e-mail to me (send a copy to your group members too). I will respond promptly.

If our group desperately needs to meet with you again, how should we set up a meeting with you?

The best way is to contact me by e-mail with a rank-ordered times that you can meet.  For example, "We would like to meet with you at one of the following times: 1., Thursday 5-7 p.m., 2. Wednesday 4-6 p.m., 3. Friday 3:30 p.m., 4. Friday 4:00 p.m., 5. Friday 4:30 p.m."  I'll send back a confirming e-mail and provide you with the first time slot on your list that's free. Don't say: "We can meet on Tuesday at 6:00" because this only leads to multiple e-mails if that slot is already taken.

Copy all of your group members e-mail addresses in your meeting request so that I can reply-to-all.

Do all of the members of our group have to be there when we meet with you?

Yes.  This may cause a one-time burden on your busy schedules, but I simply will not meet with a partial group.  In my experience, partial groups tend to be less cohesive -- and do less well on the projects overall -- than groups that are together every step of the way.  If you want my input on your project then I expect that you will find a way to coordinate each one of your schedules to meet with me.  Likewise, all members of the group should be present when you meet with Jennifer.

Do we have to run an experiment?

No. And you don't have do a survey either. There are all kinds of creative things that can be done. For example, one group built a model from which they computed the probability of contracting AIDS in the boxing ring. Another group used public records to determine the correlation between teaching and research quality for professors of different ages. Yet another videotaped interviews that they conducted with people on the street to show how people think about elementary ideas in probability.  One creative group conducted a fencing match in the classroom between a large student and his small fencing teacher.  If you're interested, I can tell you more about why this project was relevant for the class. 

Is there a specific timetable for the group project?

No. It's all up to you. In the past, the groups that have done better were those that started earlier and were more organized.

How should we word this question? Should we include a question about this? Is it OK to use business students as subjects? Where should we recruit subjects? Can we draw this conclusion if we get this result? Which statistical test should we use?

This is your project. Don't ask Jennifer or me exactly how to conduct your experiment or exactly which statistical test to use or exactly how to present your data to the class. If we tell you all of these things, it becomes our project, not yours. The educational value of such a project for you would be low.

Well then what kind of questions CAN we ask?

Examples of good questions I got last year were:

• Is the chi-square independence principle violated if the same person contributes to more than one cell even if the data are collected at different times (answer: yes).
• If we conducted a lot of t-tests on our data, is this a form of data mining in which we're likely to get a statistically significant effect by chance alone? (answer: yes).
• How do we randomize subjects to treatment groups in a way that protects us from having too many people in one condition and too few in another?
• If non-parametric tests are less powerful (i.e., less able to reject a false null hypothesis) as you said in class, then how come we didn't get statistical significance with a t-test but we did with a nonparametric test?

Can we distribute our survey or experiment in class?

No. There are several reasons for this policy. One is that, after the first day of class, you and your fellow students will have a heightened sensitivity to probabilities and statistics. This means that you are no longer representative of any subject population other than, perhaps, people who have a heightened sensitivity to statistics. But this subject population is rarely of interest. Usually, you'll want your subjects to be representative of broader populations such as "University students" or "consumers."

Even if Honors statistics students are of special interest to your project, you cannot use our class-time for your survey. Obviously, distributing a survey in class is the simplest way to collect data (fairly large class, captive subjects, etc.). But this isn't fair to your fellow classmates for numerous reasons. For example, why should your classmates spend their class time helping group after group with their surveys when they have to find creative ways of getting subjects for their own studies? The bottom line is this: no group may use class-time to collect data for their group projects.

Can we distribute our survey or experiment in another class?

I discourage you from collecting data in your other classes, particularly other BHP classes.  Surveys take time and your classmates may resent spending their time this way.  But if you insist on going this route to data collection, you must clear this with the course instructor before doing so. By "before," I do not mean 30 seconds before you plan to distribute a survey. I mean contacting the instructor days or weeks in advance and showing him/her a copy of the materials that you plan to distribute, along with a realistic estimate of the total amount of time it will take (including distribution and collection time).  Do not assume that because an instructor is "nice" that he/she will let you do this.

How large a sample do we need?

Unfortunately, there is no simple answer to this question. It has to do with the variability in your data as well as the effect size that you think will be scientifically meaningful. If your data are highly variable (e.g.,. if you expect answers to range from one extreme of the scale to the other within each cell), and if you predict that there will be a fairly large difference among cells, then you will need a very large sample. In my own work, I often use cell sizes of around 30. Cells that have fewer than 10 are often too unreliable, and cells that have more than 50 are probably unnecessary for most of your purposes. Ultimately, decisions about samples sizes are yours to make: please don't ask Jennifer or me to decide this for you.

• More Detailed Answer
  

  If you are conducting a poll and want a margin of error of 10% with 95% confidence, then you'll need 100 subjects (because 1/(sqrt N) = .10 when N=100). If you want a smaller margin of error -- which you might if you don't expect the difference between two polled groups to be more than 10% -- then you'll need to a larger N.

  If you are comparing a set of means and wish to show that they are significantly different from one another, the margin of error depends on: (a) the desired confidence level (usually 95% or 99%), (b) the N, and (c) the SD. You know this because these are the components that are used to determine the width of confidence intervals.

  The reason I can't tell you what your N should be in these situations is that the answer depends on (a) and (c). In particular, (c) is a tricky one.

  One thing you can do to get a sense of the N you'll need is to simulate a reasonable looking set of data, and test it for statistical significance. If your t values are small, try testing a larger set of data. As a general rule, you'll need at least 20 elements in each cell before you can expect the means to be stable. With more variable data, you'll need 30 or more elements. With data that show little variability, you may get by with as little as 10 or 15 per cell.

  You should be able to figure out how many cells are in your study. In a study that examines French toast eating behaviors as a function of sex and academic program, you would have 6 cells if you looked at three different programs:

  Proportion of French Toast Eaters
  	Men	Women
  Program A	1	2
  Program B	3	4
  Program C	5	6

Do we really have to collect all of that data?

If the design of your study is such that it requires, say, 200 surveys, then obviously you must collect the data. If this is too much for you to bear, then redesign your study or switch topics.

Will you (or one of the TAs) help us analyze our data?

The most challenging part of statistics is trying to figure out which type of analyses to perform in which type of situation.  The easiest part is carrying out the analyses, particularly if someone else has already told you what to do (e.g., "perform an independent samples t-test").  Your group projects require you to make some decisions about what type of analyses to perform.  In some cases, there are lots of different analyses that will work equally well.  In other cases, some analyses are superior to others.  You will have to make the call about which analyses to use on your projects.  This is your project and you should take ownership of it.  You should be able to justify the choices you made, particularly in cases where other analyses might also do the trick.  For guidance, review the textbook chapters in your reading packet, study the class notes, and click on the SPSS files that appear at the bottom of the Statistics links.  Please don't ask Jennifer, AJ, McKenzie, or me to tell you how to analyze your data. 

Should we focus more on what we did in our study or on the statistics of our study?

Focus on your study. Use statistics to bolster your conclusions, but don't show the details of your chi-square analyses or hypothesis tests. You won't impress anyone by showing a bunch of algebraic computations or hard-to-interpret computer generated output. But have such details available in case you get asked about specifics. Usually a few good graphs, tables or figures that include summaries of the statistical analyses are appropriate.

Must we give formal business presentations?

No. Feel free to present your project in any way at all (bearing in mind the evaluation criteria: originality, educational value, and overall quality).

What are the criteria that will be used for evaluating our presentations?

Your classmates, Jennifer and I will evaluate your presentations in terms of originality, educational value, and overall quality. Here is a sample Presentation Rating Form:

                         PRESENTATION RATING FORM

Which presentation of the day:     1st        2nd      3rd             

All ratings are to be made on a scale from 1 - 5 (1 = very low, 5 = very high)

1.  ORIGINALITY     

            a.  TOPIC, ANGLE                _________     

            b.  STYLE                            _________     

2.  EDUCATIONAL VALUE              _________     

3.  OVERALL QUALITY                   _________     

Comments

What does 'educational value' mean?

A project has educational value if it informs, provokes and inspires a knowledgeable listener. If your project repeats points and examples made in class, it would not have educational value for your classmates. Ask yourself whether your audience learned anything new and intellectually interesting from your proposed project. If the answer is no then you should rethink it.

Can you give an example of high and low educational values?

A high value study might be one in which people's understanding of the selection effect in roulette improves by having them play gambles repeatedly in which they lose money. A low value study is one that finds that people who spend more time studying get higher grades than those who spend more time partying.

How much time do we have?

Every group has a maximum of 15 minutes from start to finish regardless of the nature of the presentation. No group may have more time than this. I will stop your group if you attempt to don't stop within 15 minutes (I can provide a 2 minute warning if you like).  Immediately afterwards, there will be a 5 minute question period.

Can our group go first?

The fairest thing to all groups is to allow the luck of the draw to determine the order of presentation. Consequently, the groups will proceed in numerical order each day.

Can our group go on a different day?

In general, the same principle above applies. However, if all members of your group and all members of one of the other groups agree to switch days with you, you may switch.

Our group plans to use some machinery in the presentation. Is there anything in particular that we should know?

Yes.  First, in order to reduce transition time between groups, we'll use 1 laptop computer for the three groups that present in each class on a given day.  This means that Group 1 will supply the laptop computer on Wednesday (4-26), the Group 4 will supply it on the following Monday (5-1), and the Group 7 will supply it on on the following Wednesday (5-3).  Groups 2, 3, 5, 6, 8, and 9 should put their presentations on a flash drive (pen key) if they plan to use the computer.  If any group objects to this plan, let me know at least one week before your scheduled presentation.    

Second, it is your responsibility to make sure that you know how to operate any technical equipment that you wish to use. We will be on a tight time schedule on the presentations days. Come to class early enough on your presentation day that you can complete all or most of your setup (if you have any) before your group is on.  Any extra time that your group spends fidgeting around with equipment will count against your 15 minutes. The best way to guard against technical problems is to practice in our classroom several days before you're ready to present (evening is probably your best bet for finding an unoccupied classroom).  Have a backup plan ready in case of technical failure.  Let Jennifer know if you plan to use equipment other than that which is standard in the classroom.  Also, be sure that the font sizes that you show are sufficiently large and that any audio and video that you present are sufficiently loud and clear.    

What should we turn in?

As indicated in the Course Details, you should turn in one set of hard copies of your overhead transparencies, handouts, a sample of your raw data and analyses.  If you used a videotape, CD, or DVD, I would like this as well (I'll return it to you later).  Label everything you turn in with your class time and group number (e.g., 11:00 class, Group 4).  Do not turn in your transparencies (again, I just need the hard copies of those transparencies). Do not turn in a formal write-up of your project.

Keep all of your raw data (e.g., every single survey you collected) until after you have received your grade from me.  This is worth repeating: Do not throw out any raw data -- not even one survey -- until after you receive your grade from me.  I sometimes ask groups to provide me with all of their raw data.  I will lower your grade dramatically if you do not have raw data for me when I ask for it.

Do our classmates' evaluations of the presentations "count" toward our grade?

Although my evaluations determine your grade, I will take the ratings of Jennifer and your classmates into account before making my final judgments.

My group did a lousy job because one of the group members (a) was never available, (b) kept vetoing all the good ideas, (c) insisted on doing it his/her way, or (d) didn't work hard enough.

Group projects usually progress over many weeks. Groups that have these kinds of problems need to get in touch with me immediately. When these sorts of problems are called to my attention early on, I may be able to help. There's very little that I can do for a group that tells me about serious problems a few days before their presentation. There is nothing that I can do for you after the presentation.

I heard that you grade the projects pretty tough. What's up with that?

The median grade for projects for the past three years was 89%.  I don’t consider it tough when approximately half of the grades are in the A range.  However, it is true that I will assign lower grades to groups that do not put together a high value project, even if those groups put a lot of effort into their projects. 

What are the most common mistakes you see in group projects?

Some groups spend too much time on the supplementary parts of the project (e.g., video presentations), and not enough time working on the intellectual content.  Sure, video clips, costumes and humor can enhance some presentations.  But do not lose sight of the fact that this is first and foremost a science project.  You need to produce something that has real, scholarly merit.

When will you tell us our group grade?

Out of fairness to the groups that present on the later dates, I will provide detailed group project evaluations for all groups at each of the three final exam sessions.  I will bring 27 large folders to the exams, one for each of the 9 groups in all three class sections.  The folder will contain multiple written copies of my evaluation of your project and a single set of your peer evaluations.  Each of you should (a) write your name on the outside of the folder (this indicates that you received an evaluation), (b) take one copy of my evaluation, and (c) indicate if you took the packet of peer evaluations that I will include in the folder.  If you do take the peer evaluation packet, please let your other group members know, so that they can arrange a time to share these results with you.   

Do groups that present on the first day tend to get higher or lower grades than groups that present on the last day?

I try to evaluate each group on its own merits.  However, I also rank each group against the set of groups that present on a given day across the three classes (n=9 projects).  For this reason, I generally provided you with your rank on your peer evaluation forms on each evaluated dimension.  In past years, there was no difference in Group Project grades across the three presentation days.