Missing Data Analysis – Multiple Imputation, EM method

Missing Data Analysis – Multiple Imputation, EM method



This video introduces basic concept in missing data imputation including mean, regression, indication and EM method of single imputation and multiple imputation.

153 – 1391925087 – 34735

data analysis methods

usimrc

Related Posts

13 thoughts on “Missing Data Analysis – Multiple Imputation, EM method

  1. Thank you for a very comprehensive review of these very complicated subjects. So, would you recommend Mixed Effects Model over MI for a data set generated from repeated measures at a fixed time interval?

  2. Good video, but audio is really not good, very hard to hear through. Some constructive criticism for future videos, maybe wear a headset 🙂

  3. Hi, I have two time series: one is sampled monthly and the other is sampled quaterly(march, june,september and december). What is the type of missing data in this case ? MAR, MCAR or MNAR ? The missing pattern is not related to value of the variable, but related to the sampling frequency.

  4. Hey Ayumi – This is very clear and easy to follow. Whoever mentioned grammar was completely unnecessary. Follow-up Qs: 1. To perform a multiple imputation, would you use the same or different imputation methods within each dataset? 2. When you pool the regression coefficients, would you just take the mean or provide a range estimate?

  5. I do not get pooled results for my analysis e.g. regression after treating my data with multiple imputation. Do you have any suggestion about how to fix this problem. I have used questionnaires comprising of multiple items in my research.Can you share any link where step by step procedure for multiple imputation is described for multi item questionnaire?

  6. Thank you. Very informative and looking forward to reading those references.
    Your delivery would benefit from you taking advice on English language grammar.

  7. Multiple imputation is generally applicable where "multiple datasets" are created, and then each dataset is proceeded with a further analysis, e.g., regression.  Then results of regression are pooled.  So if you want to have a single set of variables, you may be referring to "single imputation".  

  8. Hi @Ayumi Shintani
    Maybe I am misunderstanding the whole concept of Multiple Imputation. I have used SPSS to perform multiple imputation for replacing my missing values. I now have the five sets of imputed data, and the original data set. Is there any way I can proceed only with the Pooled Variables?

Leave a Reply

Your email address will not be published. Required fields are marked *

Read also x