Missing Data Analysis

Missing data analysis 1

Most of us did not anticipate the problem of missing data. In most cases, there will be missing data. Therefore, Missing data should be managed and analyse. There is a YouTube video,http://www.youtube.com/watch?v=XlUdRVdT7iU, that explain about how to deal with missing data.

According to the presenter, researcher should plan for missing data – for example design the questionnaire in such a way to avoid missing data ie not applicable box. The video list these possibilities for missing data, not applicable, not available, unknown, refusal to answer and true missing (maybe accidentally)

Type of missing data
1.      Missing completely at random – probability of missing data on variable Y is unrelated to the true value of Y or other variables in the dataset
2.      Missing at random - probability of missing data on Y is unrelated to Y only after adjusting for one or more other variables
3.      Not missing at random – probability of missing data on Y is dependent on value of Y

Benefits of documenting missing data
 1.      Informs quality control reporting
2.      Allows for full disclosure in publication or presentation of data
3.      Some statistical analysis methods are dependent on missing completely at random or missing at random
4.      Useful for methodological researcher

Missing data analysis 2

Click 'Analyse'
Click 'Missing value analysis'

In 'Missing value analysis' window

Transfer ID (if you have one) to 'Case Labels'
Then, transfer all your variables to 'Quantitative variables'
Then, if applicable, transfer all variables that considered as categorical to 'Categorical variables' 
Then click 'Pattern'

In 'Missing value analysis: Patterns' window

In Display 

Tick 'Tabulated cases....'
Set to 5% in 'Omit patterns with less than .... % of cases'
Tick 'Sort variables by missing value patern'
Tick 'Cases with missing values, sorted by missing value patterns'
Tick 'Sort variables by missing value pattern
Tick 'All cases, optionally sorted by selected variable'
In Variables, select all and send to 'Additional Information for'
Click 'Continue'

In 'Missing value analysis' window

Click 'Descriptive'
Click "t Tests"
Click 'Include probabilities in table
Click 'Crosstabulations of categorical and indicator variables
Click 'Continue'

In 'Missing value analysis' window
In Estimation 

Click 'Listwise'
Click 'Pairwise'
Click 'EM'
Click 'Regression'

Then Click 'Variables' button
In Missing value analysis: Variable 
Make sure 'Use all quantitative variables' is ticked
Click 'Continue'

Click 'EM' button
In Missing value analysis: EM 
Distribution 'Normal'
Maximum iteration type '100'
Click 'Continue'

Click 'Regression' button
In Missing value analysis: Regression
Click 'Residuals'
Click 'Maximum number of predictors', then type '3'
Click 'Continue'

Finally click 'OK'

Little MCAR

Additional videos

SPSS - Missing Data Analysis [1/3]


SPSS - Missing Data Analysis [2/3]


SPSS - Missing Data Analysis [3/3]


No comments:

Post a Comment