ABENAN
SPSS produces many tables, but not every cell belongs in your report. What matters is knowing which table answers your research question, what the data allow you to say, and what they do not.
1. Start with descriptive statistics
Before any test, check the number of valid and missing cases, the mean, the standard deviation, the median where needed, and the minimum and maximum. These numbers reveal coding errors and implausible values, and they help you understand the inferential result later.
If the mean of the first group is 4.20 and the second is 4.15, a very large sample may show a significant difference, yet it remains very small in practical terms. So do not start with the p-value while ignoring what the original numbers look like.
2. What does the p-value actually mean?
The p-value measures how compatible the data are with the null-hypothesis model, under the assumptions of the test. It is not the probability that the null hypothesis is true, it is not the probability that the result happened “by chance” in any absolute sense, and it does not measure the size of the difference.
The American Statistical Association stresses that scientific decisions should not rest only on whether a single numerical threshold is crossed. So the statement “the result is important because p is below 0.05” is not enough without context, an effect size, and the quality of the design.
3. What does the 0.05 significance level mean?
The researcher sets the significance level α before the analysis, and it is usually 0.05. If the p-value is below that level, we reject the null hypothesis according to the rules of the test. If it is higher, we say the evidence is not sufficient to reject it; we do not say we have “proven there is no difference”.
The difference between 0.049 and 0.051 is not a scientific wall separating truth from error. The two values are very close, so the result should be read together with the confidence interval, the effect size, and the precision of measurement.
4. Read the confidence interval
A confidence interval gives a range of plausible values for the parameter or the estimated difference. If the confidence interval for the difference between two means does not include zero, this usually agrees with a significant result in a two-tailed test at the same level. More importantly, the width of the interval shows how precise the estimate is: a narrow interval is more precise than a wide one.
Example: a mean difference of 4 points with a confidence interval from 1 to 7 is clearer than simply writing p = 0.02. We know the direction is positive, and we see the range of the likely size.
5. Do not ignore the effect size
The effect size answers the question: “How large is the difference, or how strong is the relationship?” Examples include:
Cohen’s d when comparing two means.
Eta Squared or Partial Eta Squared in ANOVA.
The correlation coefficient r.
The Odds Ratio in logistic regression.
Do not turn the usual effect-size benchmarks into rigid verdicts. A small value can matter in a health or economic context, and a large value can matter less if the measurement is weak or the sample is biased.
6. How to interpret a t-test
In an independent-samples t-test, first check the mean and standard deviation of each group, then the test of equality of variances when you use the traditional output, and then the appropriate row for the t value. A good report includes:
The mean and standard deviation of each group.
The t value and the degrees of freedom.
The exact p-value.
The mean difference and its confidence interval.
The effect size.
Example wording: “Satisfaction was higher in the first group than in the second, with a difference of 3.2 points, t(198)=2.41, p=0.017, and a small-to-medium effect size.”
7. How to interpret ANOVA
A significant ANOVA tells you that at least two means differ, but it does not tell you which groups differ. After the overall result, you need appropriate post hoc comparisons, paying attention to the equality of variances and to the correction for multiple comparisons.
Do not write “all groups are different” just because the F value is significant. Review the post hoc comparisons, and report the direction of the differences and the effect size.
8. How to interpret chi-square
The chi-square test checks whether two categorical variables are associated. Before interpreting, check the expected frequencies. Then report the χ² value, the degrees of freedom, and p, and add a measure of the strength of the association such as Cramér’s V where appropriate.
An association does not mean that one variable causes the other, especially in cross-sectional studies.
9. How to interpret a correlation
The correlation coefficient shows direction and strength. A positive sign means the two variables tend to rise together, and a negative sign means that higher values of one go with lower values of the other. But correlation does not prove causation, and it can be influenced by a third variable or by outliers.
Look at the scatter plot instead of relying on the number alone; the relationship may be curved, or driven by a small number of cases.
10. How to interpret a regression
In linear regression, start with the model fit and R², then review the coefficients of the variables, their direction, their confidence intervals, and their p-values. Do not say that a variable “causes” the outcome unless the study design supports a causal conclusion. Also check the assumptions about the residuals, multicollinearity, and influential points.
A simple example
Research question: Are training hours related to performance scores? The results were r=0.36, p<0.001, with a confidence interval from 0.22 to 0.48.
A suitable interpretation: there is a positive, roughly moderate relationship; in this sample, the more training hours, the higher the performance score. The result is relatively precise because the confidence interval does not come close to zero. But we cannot claim that training alone caused the increase without an experimental design and control of other factors.
Common mistakes when writing up results
Writing p = 0.000; it is better to write p < 0.001.
Saying the null hypothesis was “proven” when the result is not significant.
Treating significance as evidence of a large effect.
Hiding non-significant results.
Reporting p without the means or confidence intervals.
Automatically interpreting a correlation or a regression as causation.
Running many tests without a correction or a prior plan.
A short reporting template
Write the result in this order: a description of the groups or variables, then the name of the test, then the test value, degrees of freedom, and p-value, then the effect size and confidence interval, and then an interpretation tied to the research question, without exaggeration.
For an analysis tied to your research questions and a report that explains the results clearly, see our service for survey data analysis with SPSS.
Frequently asked questions
Does a p-value above 0.05 mean there is no relationship at all?
No. It means the data did not provide enough evidence to reject the null hypothesis, given the model and the sample used. The effect may be weak, the sample may be too small, or the measurement may be imprecise. That is why the result is read together with the confidence interval and the effect size.
Should I include all the SPSS tables in my research?
No. Choose the tables that answer your research questions, and present them in clear wording instead of copying the program’s output as it is. Keep the full output for review, but the final report shows only the results that are needed.
What is the difference between statistical significance and practical significance?
Statistical significance is about the strength of the evidence against the null hypothesis, while practical significance is about the size of the effect and what it means in reality. A small difference can be significant in a large sample, and an important effect can be non-significant in a small sample.
Loading comments…