Multiple imputation in your browser

Crosstabs runs the SPSS MULTIPLE IMPUTATION approach on your device, with FCS or monotone imputation, linear regression, predictive mean matching or logistic models, and pooled descriptives and linear regression.

• Which model imputed each variable (linear regression, predictive mean matching or logistic regression), with its predictors and how many values were imputed.
• How the imputed values compare with the observed ones, per imputation and for the completed data.
• Whether the FCS chains have converged, from the iteration history of the mean and SD of the imputed values.
• Pooled means and regression coefficients with Rubin's-rules standard errors, degrees of freedom, fraction of missing information and relative efficiency.

1. 01

   Open your data and check the patterns

   Open the file in Statistics mode; running Analyze Patterns first shows whether the monotone method applies.

2. 02

   Choose the variables

   In the Missing values group, pick Impute Missing Data Values and tick at least two Variables in the imputation model. Include the variables your later analysis uses and any that predict missingness.

3. 03

   Set measurement levels

   Under Measurement level (sets the imputation model), mark each variable Scale or Categorical; categorical ones are imputed by logistic regression.

4. 04

   Choose the method and settings

   Imputation method: Automatic, FCS or Monotone. Model for scale variables: Linear regression or Predictive mean matching. Set Number of imputations (default 5), Maximum FCS iterations (default 10) and the Random seed, then press Impute.

5. 05

   Pool the analysis

   Under Analyze the imputed data (pooled results), tick Descriptives: variables and/or choose Linear regression: dependent and predictors, pick the Degrees of freedom rule, and press Run pooled analysis.

6. 06

   Download if needed

   Download imputed data (stacked CSV) gives Imputation_ = 0 for the original data and 1 to m for the completed datasets, created on this device only.

Imputation Specifications

Method actually used (and which one Automatic chose), number of imputations, model for scale variables, maximum iterations and the random seed.

Imputation Models

One row per imputed variable: model type, effects (predictors), missing values and total imputed values across imputations, in imputation order.

{variable}: Descriptive Statistics

For each imputed variable, Original Data, Imputed Values and Complete Data After Imputation, per imputation; categories are shown as counts and percents.

Iteration History

For FCS, line charts of the mean and SD of the imputed values at each iteration, one line per imputation; converged chains mix without a trend.

Descriptive Statistics (pooled)

Mean and standard error per imputation plus a Pooled row with df, 95% confidence interval, fraction of missing information, relative increase in variance and relative efficiency.

Coefficients (pooled): dependent variable {variable}

B and standard error per imputation and pooled B, SE, t, df, Sig. and 95% CI; footnotes give the joint D1 test that all slopes are 0 and the average R² (not pooled).