Prof. Dr. Larry AdamsAcademic, Author & Researcher
Research in Business and Management · Chapter 10 of 23

Quantitative Analysis with SPSS (Step by Step)

❦

SPSS (Statistical Package for the Social Sciences, by IBM) is the most widely taught analysis package in Sri Lankan business faculties. This chapter explains the main procedures and how to read the output. For fuller guidance see Pallant (2020), Field (2018), and Tabachnick and Fidell (2019). Menu names can vary slightly between versions.

Setting Up

  • Variable View: define each variable's name, type, label, values (for example 1 = Male, 2 = Female), measurement level (nominal, ordinal, scale), and missing values.
  • Data View: enter or import data from Excel (File > Open > Data).
  • Transform > Recode into Different Variables for reverse-coded items and grouping; Transform > Compute Variable for scale scores (for example, JS = MEAN(JS1 TO JS5)).
  • Data > Select Cases and Split File for analyzing subgroups.

Descriptive Statistics and Data Screening

  • Analyze > Descriptive Statistics > Frequencies for categorical variables (demographic profiles; tables and charts).
  • Analyze > Descriptive Statistics > Descriptives for means and standard deviations.
  • Analyze > Descriptive Statistics > Explore for normality (skewness, kurtosis, Shapiro-Wilk, box plots, Q-Q plots).
  • Interpreting mean scores on a 5-point scale: two methods are common: the interval approach (range 4 ÷ 5 = 0.8, so 1.00-1.80 very low, 1.81-2.60 low, 2.61-3.40 moderate, 3.41-4.20 high, 4.21-5.00 very high), or three levels (range 4 ÷ 3 = 1.33, so 1.00-2.33 low, 2.34-3.67 moderate, 3.68-5.00 high). State which you use and why.

Reliability

Analyze > Scale > Reliability Analysis; select the items of a construct; choose Cronbach's alpha, and request Scale if item deleted and item-total statistics.

AlphaInterpretation (common rule of thumb)
≥ 0.90Excellent (check for redundancy)
0.80 to 0.89Good
0.70 to 0.79Acceptable
0.60 to 0.69Questionable (sometimes accepted in exploratory work)
< 0.60Poor

Use alpha as a guide (Cronbach, 1951; Nunnally, 1978). Corrected item-total correlations below about 0.30 may indicate a poor item.

Exploratory Factor Analysis (EFA)

Analyze > Dimension Reduction > Factor: select items; use Principal Axis Factoring or Principal Components; request KMO and Bartlett's test, rotation (Varimax for uncorrelated factors, Promax or Oblimin for correlated factors), and sort by size, suppress small coefficients (for example below 0.40).

StatisticGuideline
KMO measure≥ 0.60 (above 0.80 is good)
Bartlett's testSignificant (p < .05)
EigenvaluesRetain factors above 1, and check the scree plot
LoadingsAt least about 0.40 to 0.50; avoid high cross-loadings
CommunalitiesPreferably above 0.50
Total variance explainedPreferably above 50 to 60 percent

Correlation

Analyze > Correlate > Bivariate; Pearson for interval data with roughly normal distributions; Spearman for ordinal or non-normal data. The coefficient r ranges from −1 to +1. Cohen's (1988) benchmarks: 0.10 small, 0.30 medium, 0.50 large. Always report the sample size and p-value. Correlation does not prove causation.

Group Comparisons

TestMenuUse
Independent-samples t-testAnalyze > Compare Means > Independent-Samples T TestTwo independent groups (for example male and female); check Levene's test for equal variances
Paired-samples t-testAnalyze > Compare Means > Paired-Samples T TestBefore and after
One-way ANOVAAnalyze > Compare Means > One-Way ANOVA (with post hoc, for example Tukey)Three or more groups
Chi-square test of independenceAnalyze > Descriptive Statistics > Crosstabs > Statistics > Chi-squareTwo categorical variables
Mann-Whitney U; Kruskal-WallisAnalyze > Nonparametric TestsNon-normal or ordinal data

Report effect sizes: Cohen's d (0.2 small, 0.5 medium, 0.8 large), eta squared (0.01, 0.06, 0.14), and Cramer's V.

Regression

Analyze > Regression > Linear. Put the dependent variable in "Dependent" and the independent variables in "Independent(s)"; under Statistics request collinearity diagnostics, the Durbin-Watson statistic, and R squared change; under Plots request the standardized residual plots.

Reading the output

TableWhat to look at
Model SummaryR² (proportion of variance explained) and adjusted R²
ANOVAThe F test and its significance: is the overall model significant?
CoefficientsUnstandardized B, standardized Beta (relative strength), t, p-value, and the collinearity statistics (Tolerance above 0.10; VIF below 5 to 10)
Residual diagnosticsDurbin-Watson about 1.5 to 2.5 suggests no serious autocorrelation; residual plots for linearity and constant variance

Assumptions: linearity, independence of errors, homoscedasticity, normality of residuals, and no multicollinearity. Hierarchical regression (entering blocks of variables) shows the added contribution of a block (R² change). Dummy variables are used for categorical predictors.

Mediation and Moderation

SPSS does not have these built in, but the free PROCESS macro by Andrew Hayes is widely used (Hayes, 2022). Model 4 tests simple mediation (with bootstrapped confidence intervals for the indirect effect; the effect is significant if the interval does not include zero), and Model 1 tests moderation (an interaction between X and W). The older Baron and Kenny (1986) steps are no longer recommended as the sole test; bootstrapping is preferred (Preacher & Hayes, 2008). Mean-centre variables before creating interaction terms.

Logistic Regression and Other Procedures

Use Analyze > Regression > Binary Logistic when the outcome is categorical (for example, whether a firm adopts a technology). Other procedures include MANOVA, cluster analysis, discriminant analysis, and time-series tools (Pallant, 2020).

Reporting in APA Style

Report statistics concisely, for example: "A multiple regression showed that work-life balance significantly predicted turnover intention, β = −.42, t(197) = −6.51, p < .001, and the model explained 24 percent of the variance, R² = .24, F(2, 197) = 31.2, p < .001." Present results in tables with clear titles, notes, and consistent decimals (APA, 2020).

CHAPTER 11