Mottos/Mantras/Maxims/Take-Home Lessons From Chapter 8

1. It is easy to do survey research; it is harder to do survey research correctly.
2. Do not use survey research to answer cause-effect questions. Don't ask why questions--or at least don't expect accurate answers to those questions.
3. Self-report can be misleading because participants may not know the answer (we are often strangers to ourselves),  participants may not remember the answer, or participants may not tell the truth.
4. People may not tell the truth
   because the lie--if not detected--makes them look better than the truth (social desirability bias),
   because they think the lie will support the researcher's hypothesis (obeying demand characteristics), or
   because they are not really thinking about the questions and are instead just answering a certain way (e.g., agreeing)--regardless of what the question is (i.e., they have a response set). For example, a participant might agree with every statement "yay-saying), disagree with every statement ("nay-saying"), or choose the neutral response for every statement ("central tendency bias").
5. One big challenge in survey research is specifying your population and then drawing a representative sample from that population.
6. The ideal is to get large, random samples of the population. However,
   1. Large nonrandom samples can be biased.
   2. Even large, random samples may be biased if you didn't accurately identify the population (e.g., in an election survey, you didn't correctly identify who would vote),
      you didn't have a complete list of the population (e.g., for a survey of your county, you didn't have contact information for all the people living in the county), or
      not everyone you contacted agreed to be in the survey (i.e., nonresponse bias).
7. It is important to know how you are going to make sense of your survey results before you start your survey. So,
   1. Focus your questions on your hypothesis
   2. Use fixed-alternative questions rather than open-ended questions unless you have both a system for coding those questions and a very good reason for using an open-ended question (e.g., you don't know enough about the topic to ask good fixed-alternative questions).
   3. Give your survey to friends or simulate responses to the survey and then analyze those responses to be sure you know how you are going to analyze (make sense of) your actual survey results.
8. A good survey, like a good test, should be pretested, standardized, objectively scored, and not vulnerable to response sets.
9. Edit your survey questions so that they are not long, are neither leading nor loaded, aren't double-barreled (usually avoid the words "and" or "but"), aren't negative (avoid "no" and "not"), and aren't ambiguous (avoid words that may be misunderstood).
10. When administering, collecting, analyzing, storing, and disposing of  the survey, take steps to keep participants' responses confidential. 
11. How you summarize your data depends on the type of data you have and whether you are summarizing the results from one variable or the relationship between variables.

    Type of data	Summarizing One Variable	Relationship between two variables
    Nominal	Frequency or Percentages	phi coefficient
    Ordinal	Median	Spearman's rho
    Interval or Ratio	Mean and variance or standard deviation	Pearson's r

12. The type of data will also determine what kind of inferential statistics you should use. For example, if you had interval data, you might use a t test to compare two groups, but if you had ordinal data, you should use a chi-square test instead.
13. If you do many analyses and use a .05 level of significance, the chances of making a Type 1 error (getting a significant result when the variables really aren't related) is high. For example, if you did 100 analyses using a .05 significance level, you should expect around 5 significant results--even if all your variables are unrelated.