I am conducting 2x2 repeated measure analysis for my master's thesis. I want to know how to deal with missing values in the data. 19% of the total data is missing. Can I use multiple imputation techniques to deal with missing values/EM means missing values imputation? If not what else can be done?
Secondly, if there is any way to deal with missing values, do i need to report the analysis result conducted without replacing missing values? For transparency
Thank you
Well the best of my knowledge that depends on 2 things,
1: What is the EXACT number of missing values? you have mentioned 19% (which can be a big problem if your total sample size is small). If it is large enough you can exclude those cases from the analysis and only report on cases with both measures (pre and post).
2. Secondly, what type of data are you dealing with? if it is a continuous variable you can use multiple imputations to estimate missing values.
However, it all comes down to how big a data set you have of which 19% is missing. furthermore if it is large enough you can always exclude those cases.
The answer above seems reasonable I had missing data in my masters project, so I would exclude those cases just from analysis with that missing variable the other method is to estimate missing values using for example mean of all values for those with data Personally I think it is better to exclude those with missing data, but it depends on what variables are missing, and whether some analysis can still be done on whole dataset
- Badges
- Science topic
- Similar topics
- Medicine
- Epi
- Age Groups
- Children