• When reporting nonsignificant results, it is good practice to include a power analysis (page 5). Decide what effect size would be the smallest effect size you would consider to be important, determine what power was for detecting an effect of such a size, and report that value. If that value is quite high, then you can conclude that the null hypothesis is probably close to true. If that value is not high, then a Type II error is the most reasonable explanation for the lack of statistical significance.
• Report effect size estimates (page 5). These should be reported even with nonsignificant results.
• Use confidence interval estimation (page 22). Confidence intervals for parameters of interest (including effect size estimates) are much more informative than just a value of a test statistic and its p value. In some cases the test statistic and p value may not even be necessary if you have presented appropriate confidence intervals.
• If you are going to report a p value, it is generally better to report it exactly (page 25), for example, p = .036, rather than inexactly, such as p < .05. However, three point precision is adequate (some might say four point), so if p < .001, say that, rather than p = .000027.
• When referring to the author(s), use the first person, not the third person (page 37). That is, say "I injected the subjects with the appropriate dose of Athenopram," not "the experimenter injected ....."
• Use active rather than passive voice (page 41). That is, say "We injected ...." rather than "the subjects were injected with ...." For a more extensive example, see Use of the Active Voice and the First Person in APA-Style Research Manuscripts
• Do not underline, do use italics, unless you are using a typewriter rather than a word processor (page 100), in which case you are very strange.
• While statistical symbols should be set in italic font, Greek letters, subscripts, and superscripts should not (page 140). Despite this rule, I often prefer to see the Greek letters, used as statistical symbols, in italic font.
• The title of a table should be in italic font (see pages 149 and 301).
• Consider adding to your ANOVA source tables values of eta-squared and exact p values (see example on page 162) even though the manual says "avoid columns of probability values" - they follow this statement with an example of how a source table should be prepared. The example has column of exact p values. That example also appears to have eta rather than eta-squared in one column. I recommend the use of eta-squared or omega-squared.
• Use hanging indentation, not normal indentation, in your reference list (page 299 and the examples in Chapter 4). If you can't figure out how to get your word processor to do hanging indentation, and none of your friends can either, APA will accept your reference list with normal indentation, but they will make jokes about your incompetence with respect to word processing.
• Use italics, not underlining, where appropriate in the reference list (see example in Chapter 4).
• In the reference list, for works with more than six authors, list only the first six authors. Use "et al." for the seventh and subsequent authors.
• When citing a source in the body of the manuscript, give the year upon first citation within a single paragraph but omit the year in subsequent citations to that same source in the same paragraph. I overlooked this in the Publication Manual (it is in section 3.94), but learned it from an APA-Style copy editor. The actual paragraph is reproduced here. I am not certain why the date was retained with the Hastie et al. citation but not with the Kalven & Zeisel. Because the case was simulated, an investigator might also be concerned because the mock jurors may not care about the outcome as much as they would if it were an actual case. There is, however, evidence that mock jurors show a level of involvement in simulated cases that is nearly equal that of actual jurors in real cases (Hastie, Penrod, & Pennington, 1983; Kalven & Zeisel, 1966). Another potential limitation of the external validity of the present study is that it focused on predeliberation verdicts. Participants in this study did not deliberate with each other but rather made their decisions independently. The effects of the predictor variables may differ following group deliberation, but past research has shown that a high correlation exists between predeliberation verdicts and postdeliberation verdicts (Hastie et al., 1983; Kalven & Zeisel).
   

• Wuensch's APA-Style Page

• Wuensch's Index Page

Contact Information for the Webmaster,
Dr. Karl L. Wuensch


This page most recently revised on 23. January 2007