Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • How to deal with missing values

    Hi all,

    I am currently struggling with my data. I have a quite large dataset and am going to make an index with this dataset later on. First, I need to impute missing values. How would you all do this? I am following a paper which uses mean filling, but I don't think that that is the way to go. Data can be downloaded from:
    https://datacatalog.worldbank.org/se...ataset/0037651
    unfortunately I cannot add in the data as the file is too big to insert.

    Open to any advice.
    Attached Files

  • #2
    Code:
    help mi
    is the place to start; if you are thinking about something simple you will also need to read some of the voluminous lit to see why single imputation is a problem

    Comment


    • #3
      Originally posted by Rich Goldstein View Post
      Code:
      help mi
      is the place to start; if you are thinking about something simple you will also need to read some of the voluminous lit to see why single imputation is a problem
      Yes, but at the same time, i am not sure if some of my data is missing not at random, and in that case mi is not a correct option to use.

      This is the first time I am going to impute so I am not experienced in it.

      Comment


      • #4
        even for MNAR, mi can help esp w/sensitivity analysis; a good discussion of some types of relevant sensitivity analysis is in van Buuren, S (2018), Flexible imputation of missing data, second edition, CRC Press

        Comment


        • #5
          Is it also possible to use normal mi commands in panel data sets?

          Comment


          • #6
            see
            Code:
            help mi_estimation##estimation_command
            to see what commands can be used (includes a number of estimators for panel data); actually others can sometimes be used with the "cmdok" option; see
            Code:
            help mi estimate

            Comment


            • #7
              Thank you Rich. Not sure I fully understand it still. I have also read that it is advisable to have your mi set up in wide format, which I do not fully grasp. Do you happen to know anything regarding this?

              Comment


              • #8
                it very much depends on what you want to do and what stage you are at; when I use -mi-, I use the "flong" structure (usually) so that is a long not a wide

                Comment

                Working...
                X