Reporting Statistics in APA Format

Reporting Statistics in APA Format
PSYC 330LBurnham
Reporting Results of Common Descriptive and Inferential Statistics

in APA Format
The Results sections of any manuscript (APA format or non-APA format) are intended to report the quantitative
and sometimes qualitative results of descriptive summaries and inferential tests performed to support or refute a
hypothesis. Results sections should not be discussions of the relevance of the empirical findings and should
avoid interpreting the results with respect to the hypothesis or current theories; that's what the Discussion
section is for. Thus, the Results section should avoid describing the outcome of a study as good or bad,
consistent or inconsistent with a hypothesis, unexpected or expected, and interesting or
uninteresting. Instead the Results section should present the inferential analyses and descriptive statistics in as
concise a way as possible so the reader knows exactly what evidence is being used to support or to refute a
hypothesis. It's all about the numbers!
As far as descriptive statistics, you can report them directly within the text of the Results sections (examples are
later in the document). However, if your study is complicated, such as a multilevel single-factor experimental
design, a factorial design, or a correlational study; it is often good to present your descriptive statistics as a table
or in a figure (graph). (See Chapter 5 of the APA Manual.) Remember, a picture's worth a thousand words!
When you report the results of an inferential statistic, you should report:
the test statistic (F-ratio, t-value, z-score, etc.)
the degrees of freedom (remember, there are many ways to calculate these)
the error term (if necessarynon-parametric tests and the correlations usually do not include this)
and the p-value (a priori or a posteriori alpha level: )
Importantly, statistical tests included in the results section should relate to the hypothesis being evaluated. For
example, you would not include ANOVA if your hypothesis calls for assessing a correlation between variables.
When you begin a Results section it is good to state the hypothesis, which is generally done descriptively with
words rather than symbolically (1 = 0). After your hypothesis has been stated you can indicate whether the
results support or refute the hypothesis and provide the descriptive and inferential statistics as support.
A point on formatting the statistical symbols: The values of the test statistic, means, and error values are
generally rounded to two decimal places, whereas the p-values are rounded to two or three decimal places. All
statistical symbols that are not Greek letters should be italicized (M, SD, t, r, p), Greek symbols are not (, , ).
A point on p-values. You can report the p-value as the chosen alpha level that is used for rejecting or retaining
the null hypothesis, that is, the a priori criterion for the probability of falsely rejecting a true null hypothesis,
which is generally .05 or .01. For example: t(28) = 2.99, SEM = 10.50, p < .05. Alternatively, you may report
the exact p-value that is provided after running an inferential test in a software program (SPSS, R, Minitab), that
is, the a posteriori probability that the result obtained or one more extreme is consistent with the null hypothesis.
For example: t(28) = 2.99, SEM = 10.50, p = .0057.1 If you do report the a posteriori probability and the value is
less than .001, it is customary to report p < .001. It is also acceptable to state, up front, the alpha level used as
the criterion for statistical significance in all inferential tests. If you do this you do not need to report the p-value
for each test. For example: An alpha level of .01 was used for all statistical tests.
When reporting non-significant results, the p-value is generally reported as the a posteriori probability of the
test-statistic. For example: t(28) = 1.10, SEM = 28.95, p = .268.
1
When reporting the a posterior probability you can report the p-value as being equal to the probability, as in the example here, or less than the probability when
rounded to three places. In this case, the p-value would be written as p < .006.
PSYC 330LBurnham
Examples2
Reporting a significant one-sample t-test:
The number of hours spent watching Family Guy per week (M = 11.00, SD = 2.00) was significantly
different from 10.00 hours per week, t(49) = 2.36, SEM = 0.42, p < .05.
Reporting a significant independent-samples (independent-groups) t-test:
Results show that students in the basic study group (M = 2.70, SD = 0.47) had a significantly lower GPA
at the end of their freshman year than students in the group study condition (M = 2.93, SD = 0.23), t(18) = 2.81,
SEM = 0.081, p < .05.
Reporting a significant correlated-samples (dependent groups) t-test:
The results suggested a statistically significant preference for bread with the butter side down (M = 6.75,
SD = 1.45) than bread with the butter-side up (M = 3.25, SD = 1.10), t(19) = 3.15, SEM = 1.11, p = .005.
Reporting a significant Pearson correlation:
There was a significant correlation between percent of a nation's gross domestic product spent on the
military and percentage of a nation's population that believes in the Theory of Evolution, r(31) = -0.63, p < .001.
Reporting an omnibus oneway ANOVA, with post-hoc tests:
The analysis of variances showed that the effect of group significantly influenced anxiety, F(2, 57) =
5.00, MSE = 100.25, p = .009. Post hoc analyses were conducted using Tukey's post-hoc test. Based on a
Tukey's value of CD = 2.50, the anxiety in the drug group (M = 5.25, SD = 1.80) was significantly less than in
the placebo group (M = 8.35, SD = 2.68) and the control group (M = 8.10, SD = 1.69). The anxiety in the
placebo group and the drug group did not differ significantly.
Reporting multiple a priori comparisons:
Tests of three a prior hypotheses were assessed using three independent groups t-Tests with a Bonferroni
adjusted alpha level of .016 per test (.05/3). Results suggest that the anxiety that was experienced in the drug
group (M = 5.25, SD = 1.80) was significantly less than the anxiety experienced in the placebo group (M = 8.35,
SD = 2.68), t(38) = 3.25, SEM = 0.95, p = .002, and control group (M = 8.10, SD = 1.69), t(38) = 2.72, SEM =
1.05, p = .009. However, the placebo group and the control group did not significantly differ in anxiety, t(38) =
0.65, SEM = 0.38, p = .519.
Normally, Results sections are double-spaced, the pages have 1 margins, and the section begins with the title 'Results'. For brevity, this formatting is not included
here and these examples should be viewed as reporting only the inferential statistics (not the hypotheses).
PSYC 330LBurnham
Reporting a factorial ANOVA, with non-significant interaction:

Test completion times, in minutes, were submitted to a two-way ANOVA with two levels of room noise
(noisy, quiet) and two levels of room temperature (hot, cold). The main effect of room noise was significant,
F(1, 39) = 12.65, MSE = 101.55, p = .001, suggesting that the test completion time in the noisy group (M =
3.55, SD = 1.20) was greater than the quiet group (M = 2.78, SD = 0.92). The main effect of room temperature
was significant, F(1, 39) = 5.85, MSE = 101.55, p = .02, suggesting that the test completion time in the cold
group (M = 4.04, SD = 1.25) was greater than the hot group (M = 3.58, SD = 0.99). The interaction was not
significant, F(1, 39) = 1.12, MSE = 101.55, p = .296.
Reporting a factorial ANOVA, with significant interaction:
Test error rates were submitted to a two-way ANOVA with two levels of room noise (noisy, quiet) and
two levels of room temperature (hot, cold). The main effect of room noise was significant, F(1, 39) = 8.66, MSE
= 5.65, p = .005, suggesting that number of errors in the noisy group (M = 5.75, SD = 2.98) was greater than the
quiet group (M = 3.89 SD = 1.89). The main effect of room temperature was not significant, F(1, 39) = 0.89,
MSE = 5.65, p = .351. However, the interaction was significant, F(1, 39) = 9.55, MSE = 5.65, p = .003,
suggesting that effect of room noise was greater in the hot room than in the cold room.
Reporting a chi-square test of independence:
A chi-square test of independence examined the relationship between religion (Catholic, Protestant,
Mormon, Secular/Atheist) and belief in the Theory of Evolution (Yes, No). The relationship between these
variables was statistically significant, 2 (3, N = 2000) = 9.64, p < .01. Results show the Mormons were less
likely to believe in the Theory of Evolution and Secularists/Atheists were more likely to believe in the Theory
of Evolution.
Reporting a chi-square goodness of fit test:
A chi-square goodness of fit test was used to determine whether the six Star Wars movies (Phantom
Menace, Attack of the Clones, Revenge of the Sith, New Hope, Empire Strikes Back, Return of the Jedi) were
equally preferred. Preference for the six Star Wars movies was not equally distributed, 2 (5, N = 250) = 21.36,
p < .001.

Reporting Statistics in APA Format

Uploaded by

Copyright:

Available Formats

Reporting Statistics in APA Format

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reporting Statistics in APA Format

Uploaded by

Copyright:

Available Formats

Reporting Statistics in APA Format

Reporting Results of Common Descriptive and Inferential Statistics

Reporting Statistics in APA Format

Reporting Statistics in APA Format

Reporting a factorial ANOVA, with non-significant interaction:

You might also like

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.