ritest - why might it be insignificant and how to specify

[Merve Kucuk]

Hi everyone,

I have completed the analysis of my paper on the impact of natural disasters on the savings and in the stage of finalizing my analysis using some randomization inference testing, particularly by employing the

ritest

function. My results are significant at the 1% level, however my randomization inference results are not.. Do you know why this could be the case?

I cannot share any data as I cannot copy it here since it is made available to me via a virtual platform. However I can share the results for which I have been given clearance.

Basically, I identify treatment based on the residency of the individuals in one of the treatment regions (SA4 level) at the time of the disaster.

My benchmark specification is given below:

 xtreg outcome_var treatment post did i.c_sa4_id, fe cluster(c_sa4_id) nonest

here did is simply post*treatment.

I perform the randomization inference with the below given code:

ritest treatment _b[did], reps(100): xtreg outcome_var treatment post did i.c_sa4_id, fe cluster(c_sa4_id) nonest

But my results are insignificant. I am sharing here only the results for one outcoma variable, but this reflects the situation of all the ritest results fir my other variables too.
(Pardon the low quality of the picture output, it was the only way I could export the output from the virtual desktop.)

I dont know whether it is because I am making a mistake in my specification or whether this might be caused as there is a level difference in the parallel trends of the treatment and control groups before the disaster..Or whether it is because of something else.

Any guidance and help or feedback is highly appreciated and thank you very much for your support in advance.

[Carlo Lazzaro]

Merve:
could you please share the output table of your original -xtreg,fe- code? Thanks.

[Merve Kucuk]

Hi Carlo,

I cannot unfortunately as it is in a virtual desktop and I cannot take a screen shot from it..

[Jared Greathouse]

Maybe I'm not understanding the output of ritest, but your confidence interval doesn't touch 0, for whatever that's meant to be worth.

Isn't this the same as being significant?

[Latoya Sundack]

Dear Statalist,

I have done an analysis of the impact of floods and droughts on child health.

I have created variables for floods and droughts, so my regression is

reg wasted flood drought male i.reg i,year

Given that the flood and drought variables can be considered as treatments, I am wondering if I can use randomization inference with multiple treatments in this way

ritest flood drought _b[flood] _b[drought], reps(1000) seed(123): reg wasted flood drought male i.reg i.year

Would this be correct to do it this way? When I try doing it this way, I am getting three _pm_1 values (i.e., _pm_1,_pm_2 & _pm_3. Please can you advice. Thank you.