The number of treated units in unbalanced panel

[QLgogo]

Hi,

Thanks for this great tool. I got one more question. I am not sure how 'fect' decides the number of treated units in the unbalanced panel. In my unbalanced panel, some observations of a treated unit are missing in pre-treated period (e.g., -2, -1, 0), treatment-launched period (+1 period), and post-treated period (e.g., +2, +3 period). When I try to figure out the number of treated units for each time period, I encountered the following questions:

1. I created a variable called distance (distance = current_month_index - treated_month_index) to measure the relative time to the treated month. The range of distance is [-15, 3] in my data panel at month-level. Since 'fect' view the treatment-launched month as "+1", so my relative range in 'fect' should be [-14, 4]. But the range of results got from "out.ife$est.att" is [-17, 4]. I am confused about why the lower bar is -17?
2. In the results got from "out.ife$est.att", I notice that the number of treated units for period [-17, -15] is much smaller than those for period [-14, 3], so I try to drop these units. The period [-17, -15] in 'fect' response to distance [-18, -16] in my data panel. However, I cannot drop these treated units for period [-17, -15] in 'fect' by using "drop if distance <= -16", given that "-16" is out of my distance range.
3. I tried different ways. Finally I run "drop if distance <= -13" on my data panel and successfully dropped these treated units for period [-17, -15] in 'fect'. But this dropping can also reduce the number of treated units for period [-14, 4] in 'fect'.

All these operations make me quite confused about how the number of treated units in unbalanced panel is determined in 'fect'. Could you kindly help and give some hints?

[xuyiqing]

Licheng will follow up in a few days :-) A treatment status pilot made by panelView() sent to our email address will be helpful.

Best, Yiqing

On Mon, Apr 12, 2021 at 8:48 PM QLgogo @.***> wrote:

Hi,

Thanks for this great tool. I got one more question. I am not sure how 'fect' decides the number of treated units in the unbalanced panel. In my unbalanced panel, some observations of a treated unit are missing in pre-treated period (e.g., -2, -1, 0), treatment-launched period (+1 period), and post-treated period (e.g., +2, +3 period). When I try to figure out the number of treated units for each time period, I encountered the following questions:

1. I created a variable called distance (distance = current_month_index - treated_month_index) to measure the relative time to the treated month. The range of distance is [-15, 3] in my data panel at month-level. Since 'fect' view the treatment-launched month as "+1", so my relative range in 'fect' should be [-14, 4]. But the range of results got from "out.ife$est.att" is [-17, 4]. I am confused about why the lower bar is -17?
2. In the results got from "out.ife$est.att", I notice that the number of treated units for period [-17, -15] is much smaller than those for period [-14, 3], so I try to drop these units. The period [-17, -15] in 'fect' response to distance [-18, -16] in my data panel. However, I cannot drop these treated units for period [-17, -15] in 'fect' by using "drop if distance <= -16", given that "-16" is out of my distance range.
3. I tried different ways. Finally I run "drop if distance <= -13" on my data panel and successfully dropped these treated units for period [-17, -15] in 'fect'. But this dropping can also reduce the number of treated units for period [-14, 4] in 'fect'.

All these operations make me quite confused about how the number of treated units in unbalanced panel is determined in 'fect'. Could you kindly help and give some hints?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/xuyiqing/fect/issues/15, or unsubscribe https://github.com/notifications/unsubscribe-auth/AB2PKGFOWY4EEBXCDI6YF6TTIO5KHANCNFSM422QAH6Q .

[QLgogo]

Thanks for your reply. I have figured out the reason, which is there are some missing values in my treatment dummy. I contacted ziyi liu in email and he has kindly helped. Thanks to all!