You should check this, as you otherwise risk bias. That is the short answer. No balance may indicate problems, and you may want to balance the sample. There are various ways to achieve balance.
@@SteffensClassroom Thank you~ I finally conducted a regression approach like what you do in the video, but before running regression I additionally created n-1 dummy variables for the n groups. It works!
Thanks! Very helpful. If I want to do the balance check for more than one treatment group using regression, shall I regress one by one or is there any easy way to conduct it at one time?
The clearest way would, for me, always to just regress one by one. However, nothing is stopping you from adding additional treatment dummies to your regression.
As explained. It is whether the two groups are significantly different from one another. That is, you need to check if the coefficient is significant or not.
@@josephconway16 Hi again! In general yes, as you are trying to see if there are significant difference between the groups in question. There are other measures out there in order to test if a randomization is succesful (if two groups are significantly different from each other). You could go for a non parametric test such as a rank test, supposed you feel like the assumptions of a t-test is violated. You could also go for a resampling test.
I was not expecting the Persona 5 soundtrack at the end, nice choice.
It worked out finally after following/using the same data set. Thank you very much!
Happy it helped!
how do we check balance for multiple treatment arms? do we go for a multinomial logit?
I would say pairwise with the control group :)
Thank you much but would like to know if i can get the video. Its not clear well. Struggling with kind work :(
Should my data be balanced for a DiD regression?
You should check this, as you otherwise risk bias. That is the short answer. No balance may indicate problems, and you may want to balance the sample. There are various ways to achieve balance.
Thanks! But what if I want to test across many groups at one time?
Hi. Not something I have done myself, but seems like a good place to look: www.stata.com/features/overview/pairwise-comparisons/
@@SteffensClassroom Thank you~ I finally conducted a regression approach like what you do in the video, but before running regression I additionally created n-1 dummy variables for the n groups. It works!
Great to hear thay it worked out. Thanks for the update!
Thanks! Very helpful. If I want to do the balance check for more than one treatment group using regression, shall I regress one by one or is there any easy way to conduct it at one time?
The clearest way would, for me, always to just regress one by one. However, nothing is stopping you from adding additional treatment dummies to your regression.
How does the coefficient tell me if the model is randomized or not?
As explained. It is whether the two groups are significantly different from one another. That is, you need to check if the coefficient is significant or not.
So it more focused on a p and T scores rather than how large the coefficient is?
@@josephconway16 Hi again!
In general yes, as you are trying to see if there are significant difference between the groups in question.
There are other measures out there in order to test if a randomization is succesful (if two groups are significantly different from each other). You could go for a non parametric test such as a rank test, supposed you feel like the assumptions of a t-test is violated. You could also go for a resampling test.