Generating MAR and MNAR datasets

[Nikolaos Pandis]

Hi,
I would like to generate from a full dataset with an outcome variable y and a single predictor x with 2 levels (0/1) other datasets with different proportions and patterns of missing data.
My aim is to compare the y between the 2 levels of x under different proportions and types of missingness and plot how the results vary.
I would need 2x5 datasets that contain missing values that are MAR and MNAR respectively with proportion of missing at 10%, 20%,30%,40% and 50%.
Any help to set up the code would be greatly appreciated.
Thank you,
Nikos

[Carlo Lazzaro]

Nikolaos:
https://www.guilford.com/books/Missi.../9781593853938, Chapter 3 offers a worked out example of what you're after.
That said, something along the following lines might be what you want to create with Stata:

use "C:\Program Files (x86)\Stata15\ado\base\a\auto.dta"
. g mpg_MAR=mpg

. replace mpg_MAR=. if price>=12000
(5 real changes made, 5 to missing)

. g mpg_MNAR=mpg

. replace mpg_MNAR=. if mpg_MNAR>=35
(3 real changes made, 3 to missing)

.

[Nikolaos Pandis]

Dear Carlo,
Thank you for your message and the link.
Best wishes,
Nikos