search

IALSA / longitudinal-response-pattern

0 stars 0 forks source link

2016-06-13 #1

Open andkov opened 8 years ago

andkov commented 8 years ago

Results:

Resolutions

ampiccinin commented 8 years ago

that seems generally correct! two details: 1) would be good to order them by pattern (first missing, then 0s then 1s, for example in each "column" [and can we have them in columns, so that a period takes as much space as a 0 or a 1? for ease or reading/comparison]) 2) presumably in real data there would be more 1s? and in the case of stroke, once you are a 1, you are always a 1.

andkov commented 8 years ago

Yes, These are only the first few rows. There are 319 unique patterns, but sorting may help viewing them.

1) I understand the idea: to be able to compare in columns. I may need to break it into columns to implement. We'll search around.

ampiccinin commented 8 years ago

Let's try sorting first, since it seems like that is easier. OBAS (if that is what we are looking at) has a lot of waves, so all those dots are difficult to view/count, but sorting might reduce the challenge enough that the columns would not be necessary. It would certainly be more compact!

knighttime commented 8 years ago

We are looking at MAP here. There are 18 waves of stroke data in this particular subset but we only included the first 3.

If we re-run it using the stroke_cum variable it will address the issue of "once you are a 1, you are always a 1."

knighttime commented 8 years ago

@andkov

Friday June 17th, 11am

Meeting Agenda:

  1. Change the sorting pattern in the dynamic table
  2. Finalize script for easy use
  3. Discuss how to package as a function that can be transferred easily to other data sets