1

Can someone take a shot at writing a formula using covariates into a linear regression formula? I need some feedback to see if I am on the right track. Much appreciated!

Here are my basic ingredients:

  • Dependent variable = % change in charter school enrollment
  • Independent variable = dissimilarity index score represented as a percent
  • covariate = % private school enrollment
  • covariate = school type (coded by number)
  • covariate = year
  • covariate = % special ed enrollment
  • covariate = % ELL enrollment
  • covariate = % nonwhite students
Gala
  • 8,323
  • 2
  • 28
  • 42
jeff
  • 11
  • 2
  • 5
    What is your research question; and what theory is there to help in choice of variables or point out the relevant things to check for in advance? And what do you see as the difference between an "independent variable" and a "covariate" (because there isn't any, statistically - they will all just form explanatory variables in your model in the end). What is your unit of analysis - school? district? If it is school, how can you have values both for private school enrollement and charter school? If it is district, how can you have a variable for school type? – Peter Ellis Jun 27 '13 at 05:33

1 Answers1

1

I am basically expanding Peter Ellis's comment but the name “covariate” comes from the notion that they are nuisance factors and not the main variable of interest. Often, in ANCOVA, the independent variable would be a categorical variable that was experimentally manipulated whereas the covariates are quantitative measures.

All this is merely a difference of intent and a consequence of the fact that regression, ANOVA, ANCOVA, etc. are still often discussed separately and by different people. If you look at this as a general linear model, all these variables just enter the model as predictors in similar ways, so any introductory text on multiple regression should explain how to write the formula.

See also this little explanation

Gala
  • 8,323
  • 2
  • 28
  • 42