|
|
COOKs/DFBetas for second level units |
|
Message/Author |
|
Bogdan Voicu posted on Thursday, February 11, 2016 - 10:23 am
|
|
|
Hello, I am fitting a TYPE=TWOLEVEL model with about 40,000 cases, nested in some 40 groups (countries). The question refers to saving/computing the Cook’s D and/or DFBETAs for the second level units. According to the manual, SAVEDATA: SAVE=COOKS INFLUENCE; allows saving the influence measures for each unit, but not for level 2. Is there a way to save the second level influence statistics? |
|
|
These should be saved as well with that command. These are the columns labeled OUTINFLC OUTCOOKC where C stands for Cluster level statistics. |
|
|
Thanks for the quick response! I wasn't patient enough, and I stopped the model from iterating after one hour, hoping in a short cut to level two influence stats. Today, I have let it run ... for about 2 hours, and I have got what I need. I should repeat the operation for the remaining 11 models that I have (I do not have a slow computer. For a future version, maybe it would be a good idea to allow saving only level 2 influence statistics. This would save a lot of time, since one does not need estimating the model for each of the 40.000 cases, but only for the 40 L2 units...) |
|
Back to top |
|
|