What Is a p-value and What Value Is Acceptable in SPSS
What Is a p-value
You open your SPSS output and see a column called Sig., or you come across the term p-value in a paper. These labels usually refer to the same type of information: how well your data fit the null hypothesis, H0, in a statistical test. When the p-value is smaller than the chosen significance level, usually 0,05, you have grounds to reject H0. When the p-value is greater than or equal to 0,05, you do not yet have enough statistical evidence to reject H0.
In short, the p-value answers this question: if H0 is true, how likely is it that you would observe the current result or a more extreme one. The smaller the p-value, the harder it is for the data to fit H0. A p-value is not the probability that H0 is true, and it is not the percentage effect of an independent variable on a dependent variable.
In SPSS, the p-value usually appears as Sig. or Sig. (2-tailed). In regression, you may see it in the Coefficients table under the Sig. column. In a correlation test, it usually appears under Sig. (2-tailed). For a t-test or ANOVA, the table name and location change with the procedure, so read the correct table instead of seeing one number and immediately drawing a conclusion.
The significance level is denoted by alpha and is usually set at 0,05 before you run the analysis. If p-value = 0,032, the result is smaller than 0,05. If p-value = 0,184, the result does not meet the 0,05 threshold. You can also read the posts about Statistics when you need to place the p-value in the correct analysis workflow.
Why the p-value matters in quantitative research
The p-value helps you make a decision about a statistical hypothesis, but it is only one part of the result. In a regression model, the p-value tells you whether the coefficient of a variable differs from 0 in a statistically significant way. If the p-value for variable X is smaller than 0,05, you can conclude that the data provide evidence of a relationship between X and Y in the model under consideration.
The p-value does not tell you whether the relationship is strong or weak. The regression coefficient, correlation coefficient, path coefficient, R², and effect size help you assess the direction and magnitude of the effect. A very small coefficient can still have a small p-value when the sample is large. Conversely, a coefficient that appears large can have a large p-value when the sample is small or the data are highly variable.
A p-value does not prove a causal relationship either. If your study includes a mediator, you need to read the indirect effect and its confidence interval as well. The term mediator has different interpretation criteria from those for an ordinary independent variable. If your model includes a moderator, the p-value of the interaction term is the part that requires attention. You can compare it with the posts about moderator or moderating.
In the results section, the p-value usually helps you answer three questions. First, is the hypothesis statistically supported. Second, which variables have a significant relationship with the dependent variable. Third, do the results match the direction stated in the hypothesis. You still need to check the sign of the coefficient. A small p-value with a negative coefficient does not support a hypothesis stated in a positive direction.
What p-value is acceptable
The phrase “what p value is good” can cause confusion. A p-value is not a quality score where lower is always better. Compare it with the selected alpha and report the observed value accurately when the software displays enough digits.
The table below summarises common interpretations used in quantitative theses. Each threshold has a source so that you do not write a number without methodological support.
| How to read the p-value | Decision rule | Source |
|---|---|---|
| p-value < 0,01 | Reject H0 at the 1% significance level if the test used alpha = 0,01 | (Hoàng Trọng and Chu Nguyễn Mộng Ngọc, 2008) |
| p-value < 0,05 | Reject H0 at the 5% significance level | (Hoàng Trọng and Chu Nguyễn Mộng Ngọc, 2008) |
| p-value < 0,10 | It may be treated as significant at the 10% level if the study specified alpha = 0,10 | (Field, 2013) |
| p-value ≥ 0,05 | There is not enough evidence to reject H0 at the 5% level | (Field, 2013) |
| Bootstrap confidence interval does not contain 0 | The indirect effect is statistically significant under the bootstrap procedure | (Preacher and Hayes, 2008) |
Some theses use only the rule p-value < 0,05. This is easy to present, but you need to remain consistent from the methods chapter through the results chapter. If you chose alpha = 0,05, do not call p-value = 0,078 “acceptable” in the conclusion without explaining that you are using the 10% level.
For a two-tailed test, the p-value reflects evidence in both directions. For a one-tailed test, the calculation and hypothesis must be specified in advance. You should not switch from two-tailed to one-tailed simply because the two-tailed p-value does not meet the threshold. This is the kind of mismatch a committee may ask about when the testing method does not fit the hypothesis.
How to read a p-value in the output
First identify the table, variable, and hypothesis you are reading. A Sig. value can appear in several tables, but its meaning is not exactly the same in each one. For example, Sig. (2-tailed) in a correlation table tests the significance of the correlation coefficient, while Sig. in the Coefficients table tests the regression coefficient for each variable after the other variables have entered the model.
The table below is illustrative output, not the result of a real study. The column names follow the form commonly seen in SPSS output.
| SPSS table | Variable or test | Coefficient | t / F | Sig. | Illustrative interpretation |
|---|---|---|---|---|---|
| Coefficients | Convenience → Usage intention | 0,286 | 3,412 | 0,001 | Significant at the 5% level, with a positive coefficient |
| Coefficients | Perceived cost → Usage intention | -0,074 | -1,021 | 0,309 | Not enough evidence of an effect at the 5% level |
| Correlations | Service quality and Satisfaction | 0,452 | Not applicable | 0,000 | A statistically significant positive correlation |
| ANOVA | Differences between groups | Not applicable | 4,867 | 0,009 | At least one pair of groups differs |
| Independent Samples Test | Levene's Test | Not applicable | 1,284 | 0,261 | The Equal variances assumed row can be considered |
In the Coefficients table, read the correct row for the variable and then compare the Sig. value with alpha. For the “Convenience” row, p-value = 0,001 is smaller than 0,05. The appropriate conclusion is that the regression coefficient differs from 0 in a statistically significant way in the illustrative model. You should not write that convenience “increases the result by 28,6%” until you have established whether this is a standardized coefficient and checked the measurement units of the variable.
In SPSS, a value of 0.000 in the Sig. column does not mean that the p-value is exactly 0. Write p < 0,001 when the software has rounded the value to 0,000. This is more accurate and avoids claiming that a probability is exactly zero.
When reading ANOVA, a small p-value only tells you that at least one group mean differs from another. You need to run a suitable post hoc test to identify which pairs of groups differ. For Levene's Test, the p-value is used to assess the equal-variance assumption, after which you select the appropriate row in the t-test table.
What to do when the p-value does not meet the threshold
If the p-value is greater than or equal to 0,05, first record the result exactly as it appears. Do not remove a variable, change the hypothesis, or change the test simply to make the p-value smaller. A result that is not statistically significant still needs to be reported when the analysis procedure is appropriate.
Second, check the data and the model assumptions. Review variable coding, reverse-coded items, missing values, outliers, and data entry. An outlier can affect the coefficient and p-value, but remove it only when there is a clear data-based reason. Do not remove it merely because it makes the result look worse. The post about outliers can help you review this part.
Third, check the sample size and the spread of the data. A small sample often produces unstable estimates and reduces the ability to detect a relationship. You must not add fabricated data or collect more observations from a different population simply to reduce the p-value.
Fourth, review the hypothesis form and the analytical tool. If you are testing means but the data do not meet the required conditions, you may need a suitable alternative test. The choice of alternative must be based on the research design and assumptions, not on the result you want. When the model includes a moderator or mediator, do not read the p-value of the main path instead of the interaction effect or indirect effect.
Fifth, discuss the result with your supervisor when it does not meet the threshold. In a thesis, you can report both supported and unsupported hypotheses. The committee usually wants to see whether you understand the result and have followed the analysis procedure correctly.
Distinguishing p-value, Sig., and the significance level
p-value is the statistical term. Sig. is the label SPSS uses in many tables to display the p-value. Therefore, when someone asks “what is the Sig. coefficient” while reading SPSS output, they are usually asking about the p-value in the Sig. column. The terms are directly related, but not every number labelled Sig. tests the same hypothesis.
Alpha, or the significance level, is the threshold you set in advance, usually 0,05. The p-value is calculated from the data. You compare these two values to make a decision. Do not call alpha = 0,05 a p-value, and do not write “Sig. equals 5%” when the actual output is 0,032.
The p-value is different from the regression coefficient. The coefficient shows the direction and amount of change per measurement unit or in standardized form. The p-value shows the statistical evidence that the coefficient differs from 0. The p-value is also different from R², because R² reflects the proportion of variation in the dependent variable explained by the model in a particular context.
The p-value is different from effect size as well. A result with a small p-value may still have little practical importance. When you need to assess the magnitude of an effect, examine the coefficient, confidence interval, and suitable effect-size measure. For a PLS-SEM model, for example, f² has reference levels of 0,02, 0,15, and 0,35 according to (Cohen, 1988).
If you are confusing scales, items, and the indicators used to assess a model, read what is a scale before returning to the p-value. Understanding the data structure helps you avoid using the p-value from one test to draw a conclusion about the entire scale.
Common mistakes when reading p-values
The first mistake is writing that “the smaller the p-value, the better the model.” A p-value belongs to a specific test. A model can have several paths with small p-values and still have problems with multicollinearity, reliability, or explanatory power.
The second mistake is concluding that H0 is true when the p-value is large. The safer wording is “there is not enough evidence to reject H0.” The current data do not support rejecting H0, but this does not prove that H0 is true in every context.
The third mistake is looking only at the p-value and ignoring the coefficient sign. If the hypothesis predicts a positive effect but the coefficient is -0,210 and p-value = 0,018, the result is statistically significant but does not support the positive direction of the hypothesis.
The fourth mistake is reporting Sig. = 0,000. Check the number of decimal places displayed by SPSS and write p < 0,001 when appropriate. The fifth mistake is running many tests and reporting only the result with a small p-value without describing the procedure. This reporting practice invites questions about transparency.
The final mistake is using the p-value from one test to answer the question of another test. Correlation, regression, t-test, ANOVA, mediation, and moderation each have their own purpose. Read the table name, variable name, and hypothesis before interpreting the result.
How to write this in your thesis: “The regression results show that [independent variable name] has a [positive/negative] effect on [dependent variable name], with [Beta/B] = [value] and p-value = [value]. Because the p-value is [smaller than/greater than or equal to] 0,05, hypothesis [H...] is [accepted/not supported] at the 5% significance level.” Use the correct type of coefficient, and do not describe the result as causal when the research design does not permit a causal claim.
Frequently asked questions
What is a p value in SPSS?
In SPSS, the p value usually appears under the label Sig. or Sig. (2-tailed). It is the value you compare with the significance level alpha when making a decision about the statistical hypothesis in the test you are running.
What p value is good?
If you set alpha = 0,05, a p-value smaller than 0,05 is usually treated as statistically significant. However, “good” does not mean that smaller is always better. You also need to examine the coefficient direction, effect size, confidence interval, and model fit.
Is a p-value of 0,000 equal to 0?
No. SPSS often rounds a very small value to 0,000 in the output table. When reporting the result, write p < 0,001 instead of p = 0,000 if the software display supports that conclusion.
Does a p-value greater than 0,05 mean that the hypothesis is wrong?
A p-value greater than or equal to 0,05 means that you do not have enough evidence to reject H0 at the 5% significance level. In a research hypothesis test, you can write that the hypothesis was not supported, but you should not claim that H0 is certainly true.
Why can a large coefficient have a p-value that does not meet the threshold?
A large coefficient can occur with a large p-value when the sample is small, the data are highly dispersed, the standard error is large, or the independent variables are strongly correlated. Check the sample size, outliers, model assumptions, and multicollinearity before interpreting the result.
Should you remove a variable when the p-value does not meet the threshold?
Do not remove a variable only because its p-value is large. Consider the theoretical role of the variable, the purpose of the model, the statistical assumptions, and your discussion with your supervisor. If you remove a variable, record the reason and rerun the model using a consistent procedure.
Open the exact output table you are using and record the test name, p-value, coefficient, and alpha before writing the conclusion for each hypothesis. If you need to run the analysis on your own .sav or .csv file and interpret the results according to your model, you can use M4 data analysis.