Stuff to Do
Monday's child is . . .
Ever wondered on what day you were born? Check out the calculator below.To use this calculator, input your date of birth (MM/DD/YYYY), press "OK" and then the day of the week on which you were born will appear.
Now that you know how to determine the day on which someone was born, here is an interesting piece of research. Start collecting people's date of birth. It's good to have a healthy sample (say, 50+), in fact, the more the better. Chance would dictate that there should be an even distribution of birth days throughout the week. Since there are seven days, you would expect 1/7th of your sample to appear on each day. Is that the case with your sample? Plot your data on a bar graph and then use a Chi-squared test to determine whether your observed distribution is significantly different from the expected distribution. How might you explain your results if the days of birth in your sample are not evenly distributed over the week?
Cola Wars.
Whenever the advertising types at the Pepsi Cola company run out of imagination, they bring back the Pepsi Challenge. Basically, this is a taste-test experiment to show that people prefer Pespi over its archrival Coke. You too can run your own Pepsi Challenge (or any other for that matter). Buy a couple of bottles of each type of soft drink. Place small amount in different coloured paper cups (remembering which drink went into which cup). Have your friends sample each type of drink and indicate which they prefer. (If you want to get fancy, randomly assign the "first" sample to avoid primacy effects).If there is no difference in preference, the split should be 50:50. This can be tested using a binomial test. One way to do this is to calculate an x% confidence interval around your estimate of the proportion of people who prefer Pepsi. Let p = the proportion of Pepsi people; q = the rest or 1-p. A 95% confidence interval would be p-1.96 sqrt(pq/n) and p+1.96 sqrt(pq/n). If .5 falls within that range, it is likely that people have no preference of one drink over the other.
It's probably best to use about 20 or more people in this study. What would happen, however, if the sample size were n=50 or n=100 or n=1000? Would your proportion of Pepsi preferees be more or less "statistically significant?"
Small Change Artists
A former graduate student once complained that the guys she knew were always asking for small change (coins) for vending machines. Her assertion was that women carry more loose change than do men. Here is another easily testable hypothesis on the sex-wars front. Ask a sample of male and female friends (I wouldn't ask strangers for obvious reasons) for a count of how much "coin" they have on them. Record the amounts and then do a two-sample t-test on the results (I assume this will involve a small sample). Personally, I think she just knew a bunch of cheap guys.The Deadliest Month
It is part of the conventional wisdom among Canadian health professionals that January poses a greater risk of mortality for the elderly than the milder summer months. Does this belief have an empirical basis? One quick way of checking it out is to scan the obituary pages of a major newspaper from last January and compare the mean age of the people listed with the mean age of those listed in the July obituaries. Back copies of newspapers are located (either in hardcopy or microfiche) in most libraries. Again, this looks like a standard two-sample test. It might also be interesting to examine the pattern in a major American sunbelt city. What would you expect in this instance?There are many interesting variations on this theme. For example, it is often suggested that more people die in the weeks immediately following significant holidays (e.g., Christmas) or special occasions (birthdays) than in the weeks immediately preceeding the event. Again, newpaper obituaries might be an excellent source of data for testing this proposition. Since these latter issues involve counts, it is most probable that you might want to use a Chi-squared test to assess your null hypothesis.
Got an idea for an everyday statistical question?
If you have a nagging question, why not e-mail me. If you have a doable project idea, I'll post it on this page.