I have a data set with 440 responses. I have 11 people who did not answer any question on the survey. Then there are a couple of missing values here and there outside of the full 11 non-responses. Is list wise deletion my best option? In all, I think I would end up having to delete 19 responses.
4
-
$\begingroup$ Do you mean delete the entire row, even if a person failed to answer Q41.d.iii? Or delete them only from the analysis where Q41.d.iii was used. These are both possible. It's unlikely that listwise deletion is 'best' but it might be 'most feasible given other constraints'. $\endgroup$– Jeremy MilesCommented Aug 16, 2022 at 17:16
-
$\begingroup$ For the people that didn't answer at all, I was thinking of deleting those. Then with the others that didn't answer specific questions, what would be the most appropriate method? $\endgroup$– CindyCommented Aug 17, 2022 at 15:45
-
$\begingroup$ If you do that you discard all the information someone gave you even if one trivial bit of information that you don't care about most of the time is missing. If you don't do that, your sample size is different for different analyses. Sorry, it's not easy to give straightforward answers. What sort of analyses are you planning? $\endgroup$– Jeremy MilesCommented Aug 17, 2022 at 16:36
-
$\begingroup$ Delete those, you know nothing about. For the rest: How many variables are there? If you loose information on 40 items, because one is missing, that is a pitty. $\endgroup$– BernhardCommented Aug 18, 2022 at 13:46
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
I am unsure if this question has been answered, but I will quickly provide some feedback that may help.
First, concerning the 11 participants who are missing all their data - I would exclude them from your analyses. However, if you have partial data (e.g., demographics) from these 11 participants, it may be worthwhile (and even recommended) to investigate why these individuals did not complete the survey (i.e., is it a specific subset of your sample who withdrew?).
Next, regarding the missing values outside of the missing 11 respondents. You would first need to determine the mechanism of missing data. Is this data missing completely at random (MCAR), missing at random (MAR), or missing not at random (MNAR). See this previous post I made on missing data mechanisms: Panel data analysis with a lot of missing values and a lot of variables
Once you have this mechanism determined, it will inform which approach you can take to address your missing data. For example, listwise deletion requires data to be MCAR to reduce/prevent bias in results. It may be worthwhile (and even recommended) to use multiple imputation (MI) if your data is at least MAR rather than listwise deletion. Using MI or other techniques (e.g., maximum likelihood) instead of listwise deletion may allow you to retain more of your data. Depending on your planned analyses, you may also be able to use pairwise deletion; however, like listwise deletion, this requires your data to be MCAR. This resource by Woods et al. (in press) is also fantastic for missing data and best practices/recommendations: https://psyarxiv.com/uaezh
