I have conducted a survey using a scale and out of 198 participants, 43 have not answered one of the demographic questions. I would like to check the correlation between this variable and the participants' scores on the scale. I think I need to use multiple imputation but I am not sure how!
Do you have other variables? Do you know why these values are missing? e.g., do you have a model for this?
Dear Ariana,
I think you should complete your data by surveying again (considering your sample size 198, 43 is too much).
Well, some people just did not answer the question! Most probably because they did not feel comfortable with the question.
Given that is a rather vague description I am assuming that you will not be able to treat the data either missing completely at random or missing at random after conditioning on other variables, so I would not include this variable. Imputed values will not likely get around these problems. Increasing the sample size (or getting a new sample) if done in the same way will also not change this bias.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey