Applying Remedies Step 1: Determine the Type of Missing Data Step 2: Determine the Extent of Missing Data Step 3: Diagnose the Randomness of the Missing Data Processes Step 4: Select the Imputation Method Step 1: Determine the Type of Missing Data Ignorable Missing Data Three Instances of Ignorable Specific remedies for Missing Data missing data are not Taking a sample of the needed because of the population (probability allowances for missing sampling) data are inherent in the Specific design of the data technique used. collection (a specific Operating at random condition that leads to the Explicitly accommodated missing data) in the technique used Censored Data Step 1: Determine the Type of Missing Data Missing Data Processes that Researcher should anticipate are not Ignorable the occurrence of these issues Known procedural factor Research design (i.e. errors in data entry, Data collection failure to complete the Remedies are still available whole questionnaire) for these missing data issues Unknown (i.e. refusal of the respondent to disclose sensitive data) Step 2: Determine the Extent of Missing Data Determine the patterns of the missing data To determine whether the extent or amount of missing data is low enough to not affect the results, even if it operates in a nonrandom manner.
If the extent is low apply any of the remediating approach for
missing data
How low is low enough? primary issue
Step 2: Determine the Extent of Missing Data Assessing the Extent and Deleting Individual Cases Patterns of Missing Data and/or Variables Percentage of variables with Simple remedy of deleting missing data for each case cases/variables that has a high The number of cases with missing level of missing data. data for each variable If the missing data is concentrated in a small set of cases and/or Look for nonrandom patterns variables. in data Up to the researchers if the Number of cases with no benefit of deleting the data missing data on any of the outweighs the cost of having a smaller sample size. variables Rules of Thumb 1 Missing data under 10% Sufficient number of cases with no missing data for analysis Step 3: Diagnose the Randomness of the Missing Data Processes Determine the degree of randomness present in the missing data
This will determine what remedy will best fit for the missing data
Missing at Random (MAR) and Missing Completely at Random
(MACR)
MAR requires special method of accommodating a non random component
MCAR can accommodate any type of missing data remedy
Step 3: Diagnose the Randomness of the Missing Data Processes Missing Completely At Missing at Random Random Observed value of Y truly a random sample of all Y variables There are no underlying bias to the observed data Cases with missing data and complete data cannot be distinguished from one another Step 3: Diagnose the Randomness of the Missing Data Processes Diagnostic Tests for Levels of Randomness Separating the groups between those with missing data and with complete data. Determine if there are significant differences between the two groups on other variables of interest Significant difference indicate the possibility of non random missing data process
Determining if the missing data can be classified as MCAR
Pattern of the missing data is compared with the pattern expected for a random missing data process. No significant differences are found MCAR Step 4: Select the Imputation Method