Timeline for How to run fixed-effects model on survey data in R?
Current License: CC BY-SA 4.0
19 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Feb 4 at 14:42 | comment | added | pawpaw | Sorry, I forgot to tag your name above @AlexJ | |
| Feb 2 at 21:24 | comment | added | pawpaw | Yes, I am trying to see if the coefficients on the interaction of time and other variables change and are significant. And because I have repeated observations I would like to control for time-invariant heterogeneity across individuals en.wikipedia.org/wiki/Fixed_effects_model | |
| Feb 2 at 21:16 | history | edited | pawpaw | CC BY-SA 4.0 |
added 428 characters in body
|
| Feb 2 at 21:13 | comment | added | pawpaw |
The reason I put ids = ~secu because I saw it done this way for this survey data in an example here: github.com/bwlewis/usgsd/blob/master/…. Which is probably not the best reason, but I had to start somewhere.
|
|
| Feb 2 at 3:53 | comment | added | Alex J | and specifically, what does I would like to control for individual-level heterogeneity mean? | |
| Feb 1 at 22:40 | comment | added | Alex J |
And you trying to test if the interaction of time and female, college etc. changes?
|
|
| Feb 1 at 22:39 | comment | added | Alex J |
What's secu? The most important thing for getting correct results for is passing the survey design into the svydesign object. I would expect ids = ~ PSU + SSU or something like that
|
|
| Feb 1 at 20:01 | comment | added | pawpaw | In my data primary sampling units are “metropolitan statistical areas, counties or groups of counties”. Secondary sampling units are census blocks or groups of census blocks. | |
| Feb 1 at 20:00 | comment | added | pawpaw | Hi @AlexJ, yes some individuals are observed in both time periods and some only in one (no interview or they are outside of the age range I am interested in). | |
| Feb 1 at 1:51 | comment | added | Alex J |
Re secondary sampling unit (SSU) - in survey sampling, the primary sampling unit (PSU) is the first level of clustering, and the secondary sampling unit is the level below that. For example, the PSU could be a household and the SSU a person. We expect that within a cluster, there is some correlation in responses, which is what defining ids in the survey design object accounts for when calculating standard errors etc.
|
|
| Feb 1 at 1:49 | comment | added | Alex J | In "I would like to control for individual-level heterogeneity," - does this mean you have multiple responses per individual? | |
| Feb 1 at 1:28 | comment | added | pawpaw | @AlexJ would you mind clarifying - "Can the indivual by considering a secondary sampling unit?" | |
| Feb 1 at 1:26 | comment | added | pawpaw | Hi @AlexJ, thank you for your reply. I have added more information about my variables to my question. | |
| Feb 1 at 1:23 | history | edited | pawpaw | CC BY-SA 4.0 |
added 473 characters in body
|
| Jan 30 at 23:15 | comment | added | Alex J |
Can the indivual by considering a secondary sampling unit? Then, they could be accounted for by svdesign(ids = ~ secu + individual_id)
|
|
| Jan 30 at 23:14 | comment | added | Alex J | Hi @pawpaw, welcome to CV. Can you provide some more information about what your variables mean, in the body of your question? | |
| Jan 28 at 14:55 | history | edited | kjetil b halvorsen♦ | CC BY-SA 4.0 |
added 6 characters in body
|
| S Jan 28 at 14:46 | review | First questions | |||
| Jan 28 at 14:55 | |||||
| S Jan 28 at 14:46 | history | asked | pawpaw | CC BY-SA 4.0 |