Page 341 - Statistics for Dummies

Chapter 20: Ten Tips for the Statistically Savvy Sleuth
Identify Non-Random Samples
If you’re trying to study a population but you can only study a sample of
individuals from it, how can you ensure that your sample represents the
population? The most important criteria is to select your sample in a random
fashion; that is, to take a random sample. You know a sample is random if it
had the same chance of being selected as every other possible sample of the
same size did. (It’s like pulling names from a hat.)
Many surveys aren’t based on random samples, however. For example, TV
polls asking viewers to “call us with your opinion” don’t represent random
samples. In fact they don’t represent samples at all; when you take a sample,
you select individuals from the population; for call-in polls, the individuals
select themselves.
Experiments (particularly medical studies) typically can’t involve a random
sample of individuals, for ethical reasons. You can’t call someone and say, 325
“You were chosen at random to participate in a sleep study. You’ll need to
come down to our lab tomorrow and stay there for two nights.” Such types
of experiments are conducted using subjects that volunteer to participate —
they’re not randomly selected first.
But even though you can’t randomly select the subjects (participants) for your
experiment, you can still get valid results if you incorporate the randomness in
a different way — by randomly assigning the subjects to the treatment group
and the control group. If the groups were assigned at random, they have a good
chance of being very similar, except for what treatment they received. That
way, if you do find a large enough difference in the outcomes of the groups, you
can attribute those differences to the treatment, rather than to other factors.
Before making any decisions about statistical results from a survey, look to
see how the sample of individuals was selected. If the sample wasn’t selected
randomly, take the results with a grain of salt (see Chapter 16). If you’re look-
ing at the results of an experiment, find out whether the subjects were ran-
domly assigned to the treatment and control groups; if not, ignore the results
(see Chapter 17).
Sniff Out Missing Sample Sizes
Both the quality and quantity of information is important in assessing how
accurate a statistic will be. The more good data that goes into a statistic, the
more accurate that statistic will be. The quality issue is tackled in the sec-
tion “Uncover Biased Data” earlier in this chapter. When the quality has been
established, you need to assess the accuracy of the information, and for that
you need to look at how much information was collected (that is, you have to
know the sample size).

3/25/11 8:12 PM
29_9780470911082-ch20.indd 325 3/25/11 8:12 PM
29_9780470911082-ch20.indd 325