Announcement

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

  • Update on CSDID

    Dear all
    A new version of CSDID has been made available thanks for Prof Baum.
    The main changes.
    - A bug when using notyet treated observations has been fixed. It was incorrectly estimating pretreatment effects.
    - Many people have asked how to obtain group, calendar, dynamic (event) averages. Those have also been added.
    - For the standard pretreatment test, the option "window" has been included as well. To test pretreatment ATTGT only if they fall within a window.
    Best wishes
    Fernando

  • #2
    Hey Fernando,
    I had assumed CSDID would work for repeated cross-sectional dataset and not just panel data, correct?
    However, I have an unbalanced repeated cross-sectional dataset containing firm-year observations, where each year-year there are multiple observations for individual trader-related information.
    "tab year gvar" gives the following:

    year | 0 2009 2010 2011 2012 2013 2014 2015 | Total
    -----------+----------------------------------------------------------------------------------------+----------
    2006 | 362 1,797 3,202 3,456 197 97 25 10 | 9,146
    2007 | 382 1,833 3,304 3,779 178 105 27 9 | 9,617
    2008 | 369 1,663 3,135 3,922 216 96 20 11 | 9,432
    2009 | 260 1,522 2,998 4,012 154 62 16 11 | 9,035
    2010 | 107 1,706 3,297 3,586 112 35 3 2 | 8,848
    2011 | 2 1,762 3,537 4,905 36 0 0 0 | 10,242
    2012 | 0 1,885 3,696 4,755 686 0 0 0 | 11,022
    2013 | 0 1,837 3,637 4,551 691 301 0 0 | 11,017
    2014 | 0 1,744 3,282 4,171 639 268 131 0 | 10,235
    2015 | 0 1,538 2,925 3,711 594 271 130 115 | 9,284
    -----------+----------------------------------------------------------------------------------------+----------
    Total | 1,482 17,287 33,013 40,848 3,503 1,235 352 158 | 97,878

    unfortunately it wouldn't run after I typed in:
    csdid profit, ivar(firm_id) time(year) gvar(gvar) method(dripw) notyet wboot rseed(1) agg(simple)

    The error message was "repeated time values within panel". As my data is not exactly panel because there are repeated obs in each firm-year, does that mean I can't use CSDID? Could you please take a look at it? Thank you!
    Last edited by Stata Stata; 07 Sep 2023, 22:55.

    Comment

    Working...
    X