Missing Values in Latent Class Analysis, Mixed-Mode Tree, and Mixed-Mode Cluster Analysis

Respondents with some missing data are still included in the computations used to determine the classes and are allocated to classes. Technically, this involves the assumption that the data is Missing At Random; that is, it involves the assumption that respondents with missing data are essentially identical to similar respondents with complete data, where “similar” refers to similarity on the values of the non-missing data.

The output for latent class analysis shows any filters and weights employed and shows a table with the number of responses used in the analysis for each respondent (A “response” in this context is defined as variables in the case of numeric and categorical data, and questions, in the case of Ranking and Experiment data). This table shows that most respondents have provided 25 responses (i.e., have no missing data), but 8 have no data at all. These outputs are stored in the tree nodes; to view them, right-click on a tree node and select Grow Settings and Analysis Report.

Total sample
Unweighted


Number of Responses 
    Counts
0        8
15       1
19       2
20       1
21       6
22       7
23      12
24      59
25     629

Further reading: Latent Class Analysis Software Data Analysis Software