SPSS is the most widely used statistics package in psychology, nursing, education, business and social science programs.
It is menu-driven, which makes it approachable, but it also makes it easy to run the wrong test, misread a table, or report a statistic SPSS printed for a different purpose. Most SPSS problems are not about the software at all; they are about knowing which button answers your question and which part of the output to report.
This guide walks through the full workflow: setting up the data correctly, preparing variables, running the most common analyses with their menu paths, reading the output tables that matter, and writing results in APA style.
The SPSS workflow
Define every variable in Variable View (name, label, value labels, measure), clean and recode your data, run descriptive statistics and reliability first, then choose the test that matches your question from the Analyze menu. In the output, report the key statistic, degrees of freedom, p-value and effect size, and check assumption tests such as Levene's before choosing which line of a table to read. Save your syntax so the analysis can be reproduced.
Setting up Your Data Correctly
Each row is one participant or case; each column is one variable. In Variable View, set these for every variable:
| Field | What to enter | Example |
|---|---|---|
| Name | Short, no spaces | age, gender, sat_1 |
| Label | The full question or description | Satisfaction item 1: I enjoy my work |
| Values | Codes for categories | 1 = Female, 2 = Male, 3 = Non-binary |
| Missing | Codes used for missing data | 99 = No answer |
| Measure | Nominal, ordinal or scale | Gender: nominal; age: scale |
Setting the measure correctly matters: SPSS uses it to decide which charts and options to offer, and a wrongly set measure is a common reason menus seem to be missing options.
Preparing Variables
| Task | Menu path | When |
|---|---|---|
| Reverse-code items | Transform > Recode into Different Variables | Negatively worded questionnaire items |
| Create a scale score | Transform > Compute Variable (MEAN or SUM) | Combining items into one score |
| Group a continuous variable | Transform > Visual Binning or Recode | Age bands, only if needed |
| Check for errors | Analyze > Descriptive Statistics > Frequencies | Impossible values, typos |
| Explore distributions and outliers | Analyze > Descriptive Statistics > Explore | Before parametric tests |
Always recode into a different variable rather than overwriting the original, so you can check your work.
Handling Missing Data
First, find out how much is missing: Analyze > Descriptive Statistics > Frequencies shows missing values per variable, and Analyze > Multiple Imputation > Analyze Patterns shows patterns across variables.
Small amounts of missing data (often under 5%) scattered at random are usually handled by excluding cases test by test or listwise, which SPSS does by default.
Larger amounts, or missingness linked to other variables, call for more careful methods such as multiple imputation. Whatever you do, decide it in advance where possible and report it: how many cases were missing, on which variables, and how they were handled.
Reliability Analysis
If you use a multi-item scale, check internal consistency before computing the scale score: Analyze > Scale > Reliability Analysis, select the items, and request "Scale if item deleted" under Statistics.
Report Cronbach's alpha; values around .70 or higher are usually considered acceptable. If alpha is low, check that reverse-coded items were recoded, as a forgotten recode is the most common cause.
Common Tests and Their Menu Paths
| Question | Test | SPSS path |
|---|---|---|
| Do two groups differ? | Independent samples t-test | Analyze > Compare Means > Independent-Samples T Test |
| Did scores change? | Paired samples t-test | Analyze > Compare Means > Paired-Samples T Test |
| Do 3+ groups differ? | One-way ANOVA | Analyze > Compare Means > One-Way ANOVA (add post hoc tests) |
| Are two categorical variables associated? | Chi-square | Analyze > Descriptive Statistics > Crosstabs > Statistics > Chi-square |
| Are two continuous variables related? | Pearson correlation | Analyze > Correlate > Bivariate |
| What predicts an outcome? | Linear regression | Analyze > Regression > Linear |
| What predicts a yes/no outcome? | Logistic regression | Analyze > Regression > Binary Logistic |
| Non-parametric comparisons | Mann-Whitney, Wilcoxon, Kruskal-Wallis | Analyze > Nonparametric Tests > Legacy Dialogs or Independent/Related Samples |
Checking Assumptions in SPSS
| Assumption | How to check in SPSS | If violated |
|---|---|---|
| Normality | Analyze > Descriptive Statistics > Explore > Plots: histogram, Q-Q plot, Shapiro-Wilk | Transform, or use a non-parametric test; t-tests are fairly robust in larger samples |
| Equal variances | Levene's test (printed with t-tests; option in ANOVA) | Welch's t-test or Welch's ANOVA with Games-Howell post hoc |
| Outliers | Boxplots in Explore; standardized scores via Descriptives > Save standardized values | Check for errors; report decisions to keep or remove |
| Linearity and homoscedasticity | Regression > Plots: ZRESID against ZPRED | Transform variables or use robust methods |
| Multicollinearity | Regression > Statistics > Collinearity diagnostics (VIF) | Remove or combine highly correlated predictors |
| Expected counts (chi-square) | Crosstabs > Cells > Expected | Fisher's exact test or combine categories |
Reading the Output: Which Numbers to Report
Independent Samples T-Test
SPSS shows two rows: "Equal variances assumed" and "Equal variances not assumed". Look first at Levene's test. If its significance is above .05, read the first row; if it is .05 or below, read the second (Welch's correction). Report t, df, the two-tailed p-value, and calculate or request Cohen's d (newer SPSS versions print effect sizes).
One-Way ANOVA
Report F, the two degrees of freedom (between and within groups) and p from the ANOVA table. If significant, use post hoc tests (Tukey is common when variances are equal, Games-Howell when they are not) to see which groups differ. Request eta squared through Analyze > General Linear Model > Univariate if your version does not print it.
Chi-Square
Read the "Pearson Chi-Square" row. Check the footnote about expected counts: if more than 20% of cells have expected counts below 5, use Fisher's exact test or combine categories. Report Cramér's V from the Symmetric Measures table.
Linear Regression
Four tables matter: Model Summary (R²), ANOVA (whether the model is significant overall), Coefficients (b, standardized beta, t and p for each predictor, plus VIF if you requested collinearity diagnostics), and residual plots to check assumptions.
Example: From Output to APA Write-up
Output (summarized): Levene's test F = 1.12, p = .29. Equal variances assumed row: t(118) = −2.87, p = .005. Group means: online learners M = 3.21, SD = 0.74 (n = 58); classroom learners M = 3.58, SD = 0.67 (n = 62). Cohen's d = 0.52.
Write-up: An independent samples t-test showed that classroom learners reported higher course satisfaction (M = 3.58, SD = 0.67) than online learners (M = 3.21, SD = 0.74), t(118) = −2.87, p = .005, d = 0.52. Levene's test indicated equal variances (p = .29).
The negative t simply reflects which group SPSS entered first. Report the direction in words so the reader does not have to work it out.
Example: Reporting a Multiple Regression
Output (summarized): Model Summary R² = .31. ANOVA F(3, 196) = 29.37, p < .001. Coefficients: workload b = 0.42, β = .38, p < .001; supervisor support b = −0.29, β = −.27, p < .001; years of experience b = −0.02, β = −.06, p = .37. All VIFs below 1.5.
Write-up: A multiple regression showed that workload, supervisor support and years of experience together explained 31% of the variance in burnout, F(3, 196) = 29.37, p < .001. Higher workload predicted higher burnout (β = .38, p < .001), and greater supervisor support predicted lower burnout (β = −.27, p < .001). Years of experience was not a significant predictor (β = −.06, p = .37).
Use Syntax, Even If You Use Menus
Every dialog box in SPSS has a Paste button that writes the command to a syntax window instead of running it. Paste, then run from the syntax window.
You end up with a record of exactly what you did, which you can rerun after fixing data errors, share with your supervisor, and include in an appendix if required. It also protects you when you return to the analysis weeks later and cannot remember which options you ticked.
Turning Output into APA Tables
Do not paste SPSS output tables into your paper. Rebuild them in Word with only the values you discuss: descriptive statistics by group, correlation matrices with means and standard deviations, or regression coefficients with b, SE, beta and p. Use APA table format: a bold table number, an italic title, horizontal lines only, and a note explaining abbreviations and significance markers.
Charts Worth Including
SPSS's Chart Builder (Graphs > Chart Builder) produces the visuals most assignments need: histograms to show distributions, boxplots to compare groups and spot outliers, bar charts of means with error bars, and scatterplots with a fit line for correlations.
Edit them to remove the default background shading, add clear axis labels and a title, then export as an image and add an APA figure number and caption in Word. One clear chart that supports your main finding is worth more than several default ones.
Common SPSS Mistakes
- Reading the wrong row of the t-test table after ignoring Levene's test.
- Running parametric tests on nominal variables coded as numbers.
- Forgetting to reverse-code items before computing a scale.
- Treating missing-value codes such as 99 as real data.
- Reporting "Sig. = .000" as p = 0 instead of p < .001.
- Pasting raw output into the paper.
How Our SPSS Help Works
Send your data file (.sav, Excel or CSV), questionnaire or codebook, research questions or assignment instructions, and deadline.
An SPSS analyst cleans and prepares the data if needed, runs the analyses that answer your questions, checks assumptions, and returns the output file, the syntax, APA-formatted tables and a results write-up with plain-language explanations.
Analysis is always run on your real data. Revisions within the scope of your original request are free.
Get Your SPSS Analysis Run and Explained
Send your .sav or Excel file, research questions and any instructions. An SPSS analyst runs the right tests, gives you the output and syntax, and writes the results in APA style.
Start My SPSS AnalysisFree revisions · Full refund if late · Written from scratch for your order
Frequently Asked Questions
Yes, using the data you collected. You receive the output, syntax, APA tables and a results write-up, so you can see and explain exactly what was run.
SPSS .sav files are ideal, but Excel and CSV files work too, along with your questionnaire or codebook so variables can be labeled correctly.
Check Levene's test for equality of variances. If p is above .05, read the "Equal variances assumed" row; if p is .05 or lower, read the "Equal variances not assumed" row.
The most common reason is a negatively worded item that was not reverse-coded. Check the item-total statistics and recode where needed. Very short scales also tend to have lower alpha.
Not raw output. Rebuild the key results as APA tables in your paper. Some programs ask for output in an appendix; check your guidelines.
Yes. Send the output file and your questions, and an analyst will explain which numbers to report, check your tests were appropriate, and help you write the results.
Any recent version runs the standard analyses. Newer versions print effect sizes for t-tests and ANOVA automatically; in older versions you calculate them or use the General Linear Model menus.
Data View shows the data itself, one row per case and one column per variable. Variable View is where you define each variable: its name, label, value labels, missing codes and level of measurement.
No. SPSS is for quantitative data. For interviews and other text, use qualitative software such as NVivo, ATLAS.ti or MAXQDA, or a structured coding document.