Table : TABLE 1.7Statistical Measures Often Used to Analyze Research Results
Measure
Use
Effect size
Indicates how much one variable affects another. Effect size ranges from 0 to 1: An effect size of 0.2 is called small, 0.5 moderate, and 0.8 large.
Significance
Indicates whether the results might have occurred by chance. A finding that chance would produce the results only 5 times in 100 is significant at the 0.05 level. A finding that chance would produce the results once in 100 times is significant at 0.01; once in 1,000 times is significant at 0.001.
Cost-benefit analysis
Calculates how much a particular independent variable costs versus how much it saves. This is particularly useful for analyzing public spending, such as whether investment in early education pays off in later years. (It does—see Chapter 5.)
Odds ratio
Indicates how a particular variable compares to a standard, set at 1. For example, one study found that, although less than 1 percent of all child homicides occurred at school, the odds were similar for public and private schools. The odds of such deaths occurring in high schools, however, were 18.47 times that of elementary or middle schools (set at 1.0) (MMWR, January 18, 2008).
Factor analysis
Hundreds of variables could affect any given behavior. In addition, many variables (such as family income and parental education) overlap. To take this into account, analysis reveals variables that can be clustered together to form a factor, which is a composite of many variables. For example, SES might become one factor, child personality another.
Meta-analysis
A “study of studies.” Researchers use statistical tools to synthesize the results of previous, separate studies. Then they analyze the accumulated results, using criteria that weight each study fairly. This combines studies that were too small, or too narrow, to lead to solid conclusions.