How to Use SPSS: A Guide to Running Analyses and Reading Output for Beginners
You have a .sav or .csv file and a completed questionnaire, but you do not know what order to follow when learning how to use SPSS. The problem is usually not clicking through the menus. You need to know which data to enter, which analysis fits each hypothesis, which output table to read, and how to report the result.
This guide follows the workflow a student usually needs for quantitative data: checking the file, cleaning the data, running Cronbach's Alpha, EFA, descriptive statistics, correlation, and regression. If your study includes a mediator or moderator, the analysis choice changes. You can also read the guides on mediator, moderator, and moderating.
What does it mean to use SPSS?
SPSS is software for entering, managing, and analyzing quantitative data. You enter survey data in a table, where each row usually represents one respondent and each column represents one variable. SPSS then calculates the statistics or tests you select and displays the results in the Output Viewer window.
Using SPSS correctly starts with the research question, not with opening an arbitrary menu. For example, the question “Do factors X1 and X2 affect Y?” usually leads to linear regression. The question “Is scale X reliable?” leads to Cronbach's Alpha. If you want to explore groups of factors from multiple items, you may consider EFA.
You will work mainly in two SPSS windows:
- Variable View: define variable names, data types, variable labels, coded values, and measurement levels.
- Data View: enter or check the values for each respondent.
One row of data may represent one completed questionnaire. Items such as TC1, TC2, and TC3 may measure the same construct. Standardize your variable codes from the beginning. When you read Item-Total Statistics, Rotated Component Matrix, or Coefficients, you then know what each row represents.
Why using SPSS matters in quantitative research
SPSS helps you turn raw data into evidence for each research question. The software does not choose your model, know which variable is the cause or outcome, or confirm a scale after one click on Run. You still need to rely on your model, hypotheses, and questionnaire design.
A common workflow is:
- Check missing data, data-entry errors, and unusual response patterns.
- Describe the sample using frequencies, percentages, means, and standard deviations.
- Check scale reliability with Cronbach's Alpha.
- Run EFA if the research objective and design require an exploration of factor structure.
- Create representative variables for the constructs, usually by averaging the items that meet the criteria.
- Run Pearson Correlation and linear regression to test relationships or effects.
- Check the p-value, regression coefficients, R², and assumption tests before drawing conclusions.
Keep a separate original data file, a cleaned data file, and an output folder organized by analysis round. When you remove an item, record the reason and the indicator value before and after removal. EFA output is rarely tidy the first time you run it. Removing an item, running the analysis again, and recording each round is normal as long as you have a clear methodological reason.
If you still confuse a scale with an item, read what is a scale. A scale is a set of items designed to measure a construct, while an individual column in Data View is one variable.
What SPSS values are acceptable?
There is no single SPSS number that means a result is acceptable. The threshold depends on the procedure and the purpose of the analysis. The table below uses thresholds commonly reported in quantitative theses. Each row includes a source so you can explain the choice to your supervisor instead of copying an unexplained number.
| Content in the output | Reference threshold | Interpretation | Source |
|---|---|---|---|
| Cronbach's Alpha | ≥ 0.70 | Reliability is generally considered acceptable | (Nunnally, 1978) |
| Cronbach's Alpha for an exploratory scale | ≥ 0.60 | May be acceptable in new or exploratory research | (Hair et al., 2010) |
| Corrected Item-Total Correlation | ≥ 0.30 | The item has a sufficient correlation with the total scale | (Nunnally and Bernstein, 1994) |
| KMO | ≥ 0.50 | The data have the minimum suitability for EFA | (Kaiser, 1974) |
| Bartlett's Test | Sig. < 0.05 | The variables correlate sufficiently to consider EFA | (Kaiser, 1974) |
| Eigenvalue | > 1 | A reference rule for retaining a factor | (Kaiser, 1960) |
| Total Variance Explained | ≥ 50% | The retained factors explain a sufficient reference level of variance | (Hair et al., 2010) |
| Factor loading | ≥ 0.50 | The item has a factor loading at the reference level | (Hair et al., 2010) |
| Durbin-Watson | From 1 to 3 | No serious sign of autocorrelation under the commonly used rule | (Field, 2013) |
These are reference thresholds, not permission to keep or delete items mechanically. You also need to consider item content, the difference between factor loadings, sample size, and the theoretical structure. In your thesis, state which criteria you used and which sources support them.
How to read SPSS output
After running a procedure, do not read every table in the order SPSS displays it. First identify the table that directly answers the hypothesis or analysis objective. For Cronbach's Alpha, the two important tables are Reliability Statistics and Item-Total Statistics. For EFA, you usually check KMO and Bartlett's Test, Total Variance Explained, and Rotated Component Matrix. For regression, check Model Summary, ANOVA, and Coefficients.
The table below contains illustrative output, not the result of a real study. The column names follow the way SPSS commonly displays them so you can identify where to look.
| SPSS output table | Column or row to read | Illustrative output | Interpretation |
|---|---|---|---|
| Reliability Statistics | Cronbach's Alpha | 0.812 | The scale has an Alpha above 0.70 and may be considered acceptable under the selected criterion |
| Reliability Statistics | N of Items | 4 | The scale has 4 items remaining in this analysis round |
| Item-Total Statistics | Corrected Item-Total Correlation, TC2 | 0.186 | TC2 is below 0.30, so review its content and the effect of deleting it |
| KMO and Bartlett's Test | KMO | 0.761 | KMO is above 0.50 and is suitable at the reference level |
| KMO and Bartlett's Test | Bartlett's Test Sig. | 0.000 | The value is below 0.05, so the data correlate sufficiently to consider EFA |
| Total Variance Explained | Cumulative % | 61.438 | The retained factors explain 61.438% of the variance, above 50% |
| Rotated Component Matrix | Factor loading of TC1 | 0.734 | TC1 has a factor loading above 0.50 |
| Model Summary | R Square | 0.428 | The model explains 42.8% of the variation in the dependent variable |
| Coefficients | Sig. of X1 | 0.012 | X1 has a p-value below 0.05 in the illustrative model |
A value of 0.000 in SPSS is usually reported as p < 0.001. Do not write p as exactly zero. When reading Coefficients, use Unstandardized B if you need the regression equation and Standardized Beta if you want to compare the relative effects of the variables. The Sig. column is the p-value, but read it together with the sign of the coefficient and the original hypothesis.
For example, if the Beta for X1 is 0.284 and Sig. is 0.012, you can state that X1 has a positive effect on Y in the illustrative sample and that the result is statistically significant at the 5% level. You cannot write that X1 increases Y by exactly 28.4% unless you have correctly identified the measurement unit and the type of coefficient used.
How to write this in your thesis
You can adapt the following sentence to your actual output: “The regression results show that [independent variable name] has a [positive/negative] effect on [dependent variable name], with Beta = [value] and Sig. = [value]. Since Sig. is [less than/greater than] 0.05, hypothesis [H1] is [accepted/not accepted].”
What to do when SPSS results do not meet the criteria
When an indicator does not meet the criterion, work from data checking to item review instead of immediately deleting the item with the lowest value.
First, check coding and reverse-coded items. If a negatively worded item has not been reverse-coded, Alpha or the correlation may fall sharply. For a Likert scale from 1 to 5, a reverse-coded item is often recoded using 6 - old value, but confirm that your questionnaire uses that exact score range.
Second, check missing data and unusual response patterns. A row with too many blank cells, the same response for the entire questionnaire, or an implausible completion time needs to be reviewed according to the cleaning rules stated in your methods section. You can consult the guide on outliers if you suspect that some observations are unusually different.
Third, check the Item-Total Statistics table. If the Corrected Item-Total Correlation for an item is below 0.30, read the item content and the Cronbach's Alpha if Item Deleted column. Delete an item only when there is a content or methodological basis. Repeatedly deleting items to increase Alpha while damaging the meaning of the scale creates a problem at the defence.
Fourth, if EFA does not meet the criteria, check the factor-loading matrix. An item with a low loading, a high cross-loading, or a loading on a theoretically unsuitable group may need to be removed one item at a time before running the analysis again. Save the output from every round so you can explain the process.
Fifth, consider the data and sample size. If several scales fail at the same time, the problem may come from data entry, the questionnaire, the collected sample, or the model structure. Software cannot repair a poorly designed dataset. When the cause is too few respondents or respondents who do not belong to the target population, collecting more data may be more reasonable than trying to fix the problem through technical operations.
How SPSS differs from SmartPLS and other easily confused tools
SPSS and SmartPLS both analyze quantitative data, but they organize models differently. SPSS is suitable for descriptive statistics, Cronbach's Alpha, EFA, correlation, regression, t-test, ANOVA, and PROCESS when your design supports them. SmartPLS focuses on PLS-SEM with latent variables, outer loading, AVE, CR, HTMT, VIF, path coefficient, and bootstrapping.
Do not use CFI, TLI, or RMSEA from CB-SEM to evaluate a SmartPLS model. Those indices belong to CB-SEM, and the commonly cited criteria come from (Hu and Bentler, 1999). For PLS-SEM, the process of assessing the measurement model and then the structural model, together with bootstrapping, is presented in (Hair et al., 2022).
AMOS is suitable when you use CB-SEM and need to evaluate model fit in that framework. JASP has an accessible interface for many basic analyses, while R is flexible but requires you to work with code and analysis packages. A paid SPSS service may save time on the mechanics, but you still need to understand the output because the committee may ask about every deleted item and every test choice. Running SPSS yourself usually takes more time to document each processing step, but you retain control over the data and your answers during the defence.
Common mistakes when using SPSS
Choosing the wrong variable type in Variable View. Codes from 1 to 5 for a Likert scale represent ordered response levels, not an absolutely continuous variable. Before running an analysis, check whether SPSS recognizes the variable as Numeric, whether missing values have been defined, and whether the value labels are correct.
Using spaces or special characters in variable names. Use short names such as DV1, DV2, and HL1, then write the full explanation in the Label column. This keeps the output readable and reduces errors when you use several variables in one command.
Running regression while ignoring assumptions. Check multicollinearity, residuals, outliers, and Durbin-Watson according to your research design. A good-looking p-value is not enough to show that the model is appropriate when the data have serious problems.
Misreading Sig. In the ANOVA table, Sig. indicates the significance of the overall regression model. In the Coefficients table, the Sig. value in each row indicates the significance of an individual independent variable. These values answer different questions.
Writing results before locking the data file. Each time you delete an item or filter the sample, name the file according to the processing round, such as data_clean_v2.sav. If you save only one final file, it will be difficult to answer when your supervisor asks how the initial result differed from the result after adjustment.
You can also review Statistics topics to compare commonly used procedures. If your study includes a moderator, do not use “moderation” as a decorative label. You need to identify the interaction variable and the test that fits your hypothesis.
Frequently asked questions
Where should a beginner start when learning how to use SPSS?
Start by opening the file, checking Variable View, and confirming that each row is one respondent and each column is one variable. Then clean the data, run descriptive statistics, and only after that run Cronbach's Alpha, EFA, or regression according to your research model.
Do not open too many procedures at once. Complete one step, save the output, and record your decision before moving to the next step.
Is SPSS difficult to use?
The basic menus are manageable, but choosing the correct analysis and interpreting the output correctly takes time. You may produce a table in a few minutes and still draw the wrong conclusion if you confuse Alpha, loading, Beta, and p-value.
For a thesis, learn with your own data file. Each item in the output should connect to a specific hypothesis or objective.
What should I do if SPSS shows an error and will not run?
Read the error dialog, check whether the variable has been defined as Numeric, confirm that you have not entered a string variable into a procedure that requires numeric data, and check whether the entire sample has been filtered out. For a .csv file, check the column separator, Vietnamese encoding, and header row before saving the file as .sav.
If the command runs but the output is empty, check Missing Values, Select Cases, and the filter condition. Do not delete the original data while trying to fix the error.
What Cronbach's Alpha value is acceptable in SPSS?
An Alpha of 0.70 or above is generally considered acceptable according to (Nunnally, 1978). For a new or exploratory scale, a value from 0.60 may be considered under (Hair et al., 2010). Corrected Item-Total Correlation is commonly reviewed from 0.30 according to (Nunnally and Bernstein, 1994).
You still need to read the number of items, the scale content, and Alpha if Item Deleted. A high Alpha caused by duplicate items does not automatically prove the validity of the scale.
Do I need to run every test in SPSS?
Run only the analyses that answer your model and research questions. Adding t-test, ANOVA, correlation, or EFA simply to create more tables will make the results chapter longer without improving its logic.
Create a small table linking each hypothesis to its variables, procedure, output table, and decision criterion. This shows you what to run before opening SPSS.
Return to your data file now, check Variable View, save a copy of the original data, and list the analyses required by each hypothesis. If you need to run analyses on your own .sav or .csv file and track each step, DoThesis M4 analysis is the module for that work.