Stata Experts:
I have been struggling for several days with my data analysis plan and I am hoping for some clarity.
1. I am working with ECLSK:2011, a nationally representative dataset which requires the use of the .svyset command. (data is nested --
children in schools, schools in districts (the primary sampling units). I have done that and all is fine with respect to descriptive analysis.
2. I am looking at the effect of maternal education (5-level categorical predictor based on educational attainment: below high school, high school, some college, etc) ) on children's math achievement (represented by a z-scored continuous variable) in the fall of kindergarten.
3. I would like to analyze the data using multilevel modeling with fixed effect to control for school effects.
4. This is a four level model:
Step 1: Regress maternal characteristics on math achievement:
... maternal characteristics include maternal age, poverty level, work status, occupational prestige
Step 2: Add maternal activities at home; i.e. frequencies of activities like building, working with numbers, playing with puzzles. The frequencies include: never, once or twice/week, three to six times/weeks and everyday. I used PCA to analyze these activities in the form of PC_1 and PC_2.
Step 3. Add maternal expectations for children's educational attainment (also a 5-level categorical variable: below high school to professional degree); and finally
Step 4. Add covariates found in research to be associated with math achievement, i.e. like single parent hh, poverty level, number of siblings, child age.
In Stata, the svy prefix cannot be used with: .xtreg fe. Is there a different way to do this analysis?
For example, what about just using the cluster option with OLS regression? cluster(school_id)
Is .mixed a possibility? I am not measuring Level 2 data and I am not an expert in this type of analysis. Would .areg work by absorbing (school_id)?
Any help would be gratefully accepted.
Best,
Pat
I have been struggling for several days with my data analysis plan and I am hoping for some clarity.
1. I am working with ECLSK:2011, a nationally representative dataset which requires the use of the .svyset command. (data is nested --
children in schools, schools in districts (the primary sampling units). I have done that and all is fine with respect to descriptive analysis.
2. I am looking at the effect of maternal education (5-level categorical predictor based on educational attainment: below high school, high school, some college, etc) ) on children's math achievement (represented by a z-scored continuous variable) in the fall of kindergarten.
3. I would like to analyze the data using multilevel modeling with fixed effect to control for school effects.
4. This is a four level model:
Step 1: Regress maternal characteristics on math achievement:
... maternal characteristics include maternal age, poverty level, work status, occupational prestige
Step 2: Add maternal activities at home; i.e. frequencies of activities like building, working with numbers, playing with puzzles. The frequencies include: never, once or twice/week, three to six times/weeks and everyday. I used PCA to analyze these activities in the form of PC_1 and PC_2.
Step 3. Add maternal expectations for children's educational attainment (also a 5-level categorical variable: below high school to professional degree); and finally
Step 4. Add covariates found in research to be associated with math achievement, i.e. like single parent hh, poverty level, number of siblings, child age.
In Stata, the svy prefix cannot be used with: .xtreg fe. Is there a different way to do this analysis?
For example, what about just using the cluster option with OLS regression? cluster(school_id)
Is .mixed a possibility? I am not measuring Level 2 data and I am not an expert in this type of analysis. Would .areg work by absorbing (school_id)?
Any help would be gratefully accepted.
Best,
Pat
Comment