None of the methods that you mention would be recommended because they imply very restrictive assumptions about the missing data mechanism, result in a loss of statistical power, and/or can lead to biased results. A better way to handle missing data is to use modern statistical approaches such as full information maximum likelihood estimation (FIML) and multiple imputation (MI). These techniques can now be applied easily to both simple types of analyses (e.g., regression or ANOVA) and complex models (e.g., path analysis, factor analysis, structural equation modeling) in many statistical software programs.

FIML and MI help researchers make use of all available data (including observations with missing values), retain as much statistical power as possible, and minimize bias (e.g., by including auxiliary variables). FIML and MI rest on less restrictive assumptions about the missing data mechanism (i.e., missing at random or MAR data) than “traditional” methods such as listwise deletion, which imply the more restrictive assumption of missing completely at random (MCAR) data. Using FIML or MI thus tends to be less problematic than using methods such as listwise or pairwise deletion.

See

Enders, C. K. (in press). Missing data: An update on the state of the art. Psychological Methods.

Enders, C. K. (2022). Applied missing data analysis (2nd ed.). New York: Guilford Press.

Schafer, J. L., & Graham, J. W. (2002). Missing data: Our view of the state of the art. Psychological Methods, 7(2), 147–177. https://doi.org/10.1037/1082-989X.7.2.147