Hot Deck

TL;DR

Fills missing values by using observed values from similar records (donor cases).
Applicable to both continuous and categorical variables and can preserve relationships between variables.
Simple to implement but can be time-consuming and its accuracy depends on the similarity and representativeness of available cases.

Definition

Hot deck imputation is a method of handling missing data in statistical analysis. It is based on the idea of using available data from similar cases to infer the missing values in a given case.

Explanation

Hot deck imputation fills missing values by locating other cases in the dataset that are similar on relevant characteristics and using their observed values as estimates for the missing entries. It can be applied to a wide range of data types (continuous and categorical) and can preserve relationships between variables, which is important for accurate statistical analysis.

Advantages described in the source:

Relatively simple and easy to implement.
Works with both continuous and categorical variables.
Can preserve inter-variable relationships.
Can improve precision and accuracy by using multiple donor points, which may help with rare or unusual data.

Limitations and potential drawbacks described in the source:

Accuracy of imputed values depends on the quality and similarity of available donor cases; non-representative donors can produce biased or unreliable estimates.
Can be time-consuming and labor-intensive because it may require manual searching and selection of similar cases.
Scalability can be limited when applied to large or complex datasets.

Examples

Example 1: Household income survey

Imagine that you are conducting a survey on the income levels of households in a certain area. Some of the households refuse to disclose their income levels, so you are left with missing data. To handle this missing data, you can use hot deck imputation by finding households with similar characteristics (e.g. number of people in the household, education level of the head of the household, etc.) and using the income levels of those households as estimates for the missing values.

Example 2: Age and blood pressure study

Imagine that you are studying the relationship between age and blood pressure in a group of individuals. Some of the individuals do not want to disclose their blood pressure levels, so you are left with missing data. To handle this missing data, you can use hot deck imputation by finding individuals with similar ages and using the blood pressure levels of those individuals as estimates for the missing values.

Notes or pitfalls

Imputed values may be biased or unreliable if donor cases are not representative of the missing cases.
Manual selection of similar cases can be labor-intensive, limiting the method’s practicality on large datasets.
The method’s effectiveness depends on having a sufficient number of appropriate donor cases.

Missing data
Imputation