How to Handle Missing Data in SPSS
You’ve spent weeks collecting data—distributing questionnaires, conducting interviews, and entering responses into SPSS. Then you run your first frequency table and notice the dreaded dots. Missing data.
This is one of the most common frustrations for Kenyan students conducting research. Whether you’re an undergraduate at the University of Nairobi, a master’s student at Kenyatta University, or a PhD candidate at Moi University, missing data is almost inevitable. But here’s the good news: ignoring missing data is not your only option, and it’s certainly not the best one.
At Proposal Writers Kenya, we help students navigate every stage of their research journey—from proposal writing to data analysis. In this guide, we’ll walk you through everything you need to know about handling missing data in SPSS, from identifying missing values to applying the right imputation techniques for your thesis or dissertation.
What Is Missing Data and Why Does It Matter?
Missing data occurs when information for one or more variables is absent from your dataset. In SPSS, missing values typically appear as dots (system-missing) or as user-defined values like -99.
Why is missing data a problem?
When you run analyses in SPSS, most procedures simply omit missing values . This means:
Your sample size effectively shrinks with every analysis
You lose statistical power—harder to find significant results
Your estimates may become biased, especially if missingness is not random
Standard errors and confidence intervals become unreliable
The key insight is this: Most SPSS procedures assume a complete dataset. Ignoring missing values leads to results that may not accurately represent your population.
Understanding Types of Missing Data (MCAR, MAR, MNAR)
Before choosing a handling method, you need to understand why data is missing. This helps you select the right strategy and defend it in your methodology chapter.
MCAR (Missing Completely at Random)
The probability of missingness is unrelated to any observed or unobserved data. Example: Equipment failure randomly affects some trials.
MAR (Missing at Random)
Missingness depends on other observed variables in your dataset. Example: Older participants are more likely to skip a question about internet usage.
MNAR (Missing Not at Random)
Missingness depends on the value itself or unobserved data. Example: High-income respondents refusing to report their income.
Why does this matter? The type of missingness determines which handling method is appropriate. If your data is MNAR, even advanced techniques like multiple imputation may not fully solve the problem—you may need additional data collection or sensitivity analysis.
Step 1: Identifying Missing Values in Your Dataset
SPSS recognizes two types of missing values :
System-Missing Values
These are automatically recognized by SPSS and displayed as a dot (.) in the Data View. They occur when cells are left blank in your data file. System-missing values cannot be included in most analyses .
User-Defined Missing Values
These are numeric values you define as missing—commonly -9, -99, or 999. These values should not exist elsewhere in your data.
How to define user-missing values in SPSS:
Go to Variable View tab
Click the Missing cell for the variable you want to define
Click the button on the right side to open the Missing Values dialog box
Select Discrete missing values and enter your chosen value (e.g., -99)
Click OK
For string variables, SPSS does not automatically recognize empty strings as missing—you must define them as user-missing
Step 2: Exploring the Pattern of Missing Data
Before handling missing data, explore its patterns. IBM SPSS Missing Values offers diagnostic reports to help you understand missingness in your dataset .
Key Reports to Run:
Data patterns table: A case-by-case overview of your data showing which variables are missing for each case
Univariate statistics: Count and percentage of missing values for each variable
Little’s MCAR test: Determines if your data is MCAR—essential for selecting appropriate methods
What to Look For:
The percentage of missing data per variable (generally, <5% is manageable)
Whether missingness patterns are random or systematic
Extreme values that may indicate data entry errors
If you have a serious missing data problem—say, more than 20% missing in key variables—consider whether you can re-collect data or whether your sample is still representative.
Step 3: Choosing the Right Handling Method
A. Listwise Deletion (Default in Many SPSS Procedures)
What it does: Removes any case with any missing value from the analysis .
When to use: Low missing rates (<5%) and MCAR data.
Pros: Simple, preserves data integrity.
Cons: Reduces sample size dramatically, can bias results if data is not MCAR.
SPSS default: Many procedures (regression, correlations) use listwise deletion by default. The CORRELATIONS command, for example, uses pairwise deletion—including cases where data is available for each pair .
B. Pairwise Deletion
What it does: Uses all available data for each analysis. For correlations, SPSS includes all cases with valid data for each specific pair of variables .
When to use: When you have different missing patterns across variables and want to maximize data usage.
Example: In SPSS correlations, the CORR command uses pairwise deletion by default. If your dataset has 100 cases but only 80 have data for both variables X and Y, the correlation between X and Y is based on 80 cases—not 100 .
C. Mean Imputation (Transform > Replace Missing Values)
What it does: Replaces missing values with the variable mean, median, or nearby points .
Available methods in SPSS:
Series mean: Overall mean of the variable
Mean of nearby points: Average of valid surrounding points within a set span
Median of nearby points: Median of valid nearby points
Linear interpolation: Estimates using surrounding valid points
Linear trend at point: Predicted values from linear trend regression
When to use: Quick fix for small amounts of missing data.
Important warning: Mean imputation is generally bad practice. It reduces variance, distorts correlations, and can bias your results. Many statisticians recommend against it .
D. Subject-Specific Mean (For Scale Variables)
What it does: Computes a mean across multiple scale items for each case .
SPSS function: MEAN.n(variable1, variable2, ...)
Example: MEAN.4(X1, X2, X3, X4, X5) calculates the mean only if at least 4 variables are observed .
When to use: When averaging Likert scale items for a respondent.
Pros: Uses the individual’s pattern rather than the variable mean.
Cons: Not appropriate for single-variable imputation.
E. Multiple Imputation (Most Robust Method)
What it does: Creates several complete datasets with different replacement values, analyzes each, and pools results .
Where to find it: Analyze > Multiple Imputation > Impute Missing Data Values .
Methods: Fully Conditional Specification (FCS) with Predictive Mean Matching (PMM) for scale variables .
When to use: MAR or MCAR data with >5% missingness.
Pros: Preserves data structure, accounts for uncertainty, handles complex missing data patterns .
Cons: More complex, requires careful setup, may not work well with very large missing percentages (>30%).
Step 4: Performing Multiple Imputation in SPSS
Step-by-step process:
Go to Analyze > Multiple Imputation > Impute Missing Data Values
In the Variables tab, add variables with missing values to the “Impute” box and add predictors to the “Predictors” box
Set a name for your imputed datasets (e.g.,
imp_)In the Method tab, choose Fully Conditional Specification (FCS)
For each scale variable, select Predictive Mean Matching (PMM)
In the Output tab, set the number of imputations (typically 5) and iterations (e.g., 10)
Click OK
What to expect: SPSS creates multiple completed datasets that you can view in the Data Editor. A new variable, Imputation_, indicates which imputation each case belongs to (0 for original data, 1–n for imputed values)
Step 5: Interpreting SPSS Output for Multiple Imputation
After imputation, your output will include:
Imputation Summary Table: Shows variables imputed, number of missing values, and method used
Iteration History Table: Displays changes in variable means over iterations—confirms convergence
Descriptive Statistics Table: Reports mean and standard deviation for each imputed variable across datasets
When running analyses after imputation, use procedures that support pooled results (e.g., REGRESSION). SPSS will automatically produce parameter estimates that account for variation within and between imputations
How to Report Handling Missing Data in Your Thesis
Your methodology chapter must transparently describe how you handled missing data. Include:
Method used: What technique(s) you applied
Why: Justification based on missing data type (MCAR/MAR/MNAR)
Extent: Percentage of missing data per variable
Impact: How the method preserves or affects your sample
Example reporting language:
“Missing data were identified using SPSS Missing Value Analysis. Little’s MCAR test indicated data were missing completely at random (χ² = 12.34, p = .234). Variables with less than 5% missing data were handled using listwise deletion, while variables with 5-15% missing values were imputed using multiple imputation with predictive mean matching (5 imputations, 10 iterations).”
Common Mistakes to Avoid
| Mistake | Why It’s a Problem |
|---|---|
| Ignoring missing data | Your results may be biased and statistically weak |
| Using mean imputation as default | Reduces variance, distorts correlations |
| Not distinguishing system-missing from user-missing | May incorrectly treat valid values as missing |
| Using listwise deletion when data is MAR | Introduces bias, reduces sample size |
| Failing to report missing data handling | Your methodology chapter is incomplete |
Practical Tips for Kenyan Students
Before collecting data: Design questionnaires to minimize missing responses—test your instrument thoroughly
During data entry: Use code -9 or -99 consistently for missing responses
Always check: Run frequencies and descriptives first to identify missing patterns
Document everything: Keep a clear record of how you handled missing data for your thesis
Consult your supervisor: Discuss your missing data strategy before analysis—your supervisor may have preferences
Conclusion
Missing data is a common challenge in research, but it’s not a roadblock. By understanding the type of missingness and applying the right handling method, you can ensure your analysis remains robust and your results credible.
The gold standard for handling missing data is multiple imputation—it preserves data structure, accounts for uncertainty, and produces statistically defensible results . However, simpler methods like listwise deletion may be appropriate when missingness is minimal (less than 5%) and MCAR.
Whatever method you choose, transparency is key. Document your approach, justify your choices, and consult your supervisor. This demonstrates rigor in your research—something examiners notice.
Need help with your data analysis or thesis proposal? At Proposal Writers Kenya, we provide expert SPSS support for Kenyan students at all academic levels. From data entry to multiple imputation, our team can help you get accurate results that impress your supervisors. Get your free quote today and let’s get your research on track.