A Question about Analyzing Shool-nested Panel Data

[Minda Tan]

Hi, I have a question about analyzing school-nested panel data. I am using education panel data to examine the relationship between students' test scores and school SES. The school SES is generated by averaging the family SES of students at this school. The dataset has two waves, including the 2014 wave and the 2015 wave. Given that students' family SES cannot change a lot within one year, I think the fixed-effect model doesn't fit my situation. How should I investigate this dataset properly with Stata?

My current codes in Stata:

xtmixed logstts schses stses stexpect i.wave || schids: vce(r)

"logstts" is the log of student's test scores. "stses" is student's family SES. "stexpect" is student's educational expectations.

I'm also wondering whether using the random-effect model can be helpful? Like:

mi xtset ids wave

xtreg logstts schses stses stexpect, re vce(cl schids)

"ids" is student's unique id in this dataset.

Thank you!

[Carlo Lazzaro]

Minda:
with two waves of data ony, I suspect that -re- will not perform well either.
Actually, as students re obviously nested withun school, -xtmixed- (that was superseeded by -mixed- from Stata 13 onwards) is the way to go.
You can also consider -xtreg, re mle-.

[Minda Tan]

Minda:
with two waves of data ony, I suspect that -re- will not perform well either.
Actually, as students re obviously nested withun school, -xtmixed- (that was superseeded by -mixed- from Stata 13 onwards) is the way to go.
You can also consider -xtreg, re mle-.

Hi Carlo! Thank you for the helpful advice. I will try it immediately!

[Minda Tan]

Hi Carlo. I have another quick question. Do you think "Linear regression with multiple fixed effects" can solve my problem?

reghdfe logstts schses stses stexpect i.wave, absorb(schids)

[Carlo Lazzaro]

Minda:
it may be an option if you're interested in within-panel variation only.
That said, please note that -mixed- is more similar to -re-.

[Minda Tan]

Minda:
it may be an option if you're interested in within-panel variation only.
That said, please note that -mixed- is more similar to -re-.

That's what I want! Thank you so much!