The main objective of this project is to evaluate approaches to imputation and make use of them on data accumulated by Equal Employment Opportunity Commission. Initially, I talk about a number of imputation methods and examine theory of multiple imputation (MI). After that, I evaluate issues with missing data and describe an artificial data simulation. I explain simulation according to EEOC dataset listing figures of workers by ethnicity in large establishments. Mean imputation and MI are used on simulated datasets. In the initial situation, we impute data for nonresponding establishments. The higher we impute, the more our resulting population means. In the next situation, we simulate item nonresponse. I have discovered mean imputation and MI produce equivalent means. The means aren’t impacted by % of missingness irrespective of imputation approaches. The outcomes suggest MI produces larger standard error than mean imputation. Last the % of missingness doesn’t have any influence on standard error in case of MI.
Contents
Chapter 1: Introduction
Chapter 2: Classical Methods Dealing with Missing Values
2.1 Historical Development of Treatment
2.2 Overview of Single Imputation
2.3 Listwise Deletion
2.4 Pairwise Deletion
2.5 Some Single Imputation Methods for Dealing with Missing Data
2.5.1 Mean Imputation
2.5.2 Regression Imputation
2.5.3 Hot Deck Imputation
Chapter 3: Multiple Imputation
3.1 History Development
3.2 Model Assumption
3.2.1 Data Model
3.2.2 Prior Distribution
3.2.3 The Nonresponse Mechanism
3.3 General Idea of Multiple Imputation
3.4 Advantages of Multiple Imputation
3.5 Concepts of Multiple Imputation
3.6 Key Features of Multiple Imputation
3.7 The Methods in the SAS Proc MI Procedure
Chapter 4: Artificial Data Simulation
4.1 Simulation Description
4.2 Outputs
Chapter 5: Equal Employment Opportunity Data
5.1 Structure of EEO-1 Data
5.2 Discussion of Missing Data Mechanism
5.3 EEOC’s Method for Handling Missing Value
Chapter 6: Simulations Based on EEO-1Data
6.1 SIC Code
6.2 Simulation Description and Outputs
6.2.1 First Scenario: Nonrespondents in 2003
6.2.2 Second Scenarios: Simulated Item Nonresponse…..
Source: UMD