0

I have a group of repeated measurement data, the dependent variable (y) was skewed distribution, when I build the GEE, should I transform y to a normal distribution variable first? Or Can I build the GEE directly without any connection functions?

My R program was like:

geeglm(y ~ Times, data=GEEData, id=id, family = gaussian, corstr = "exchangeable")

Lookout, the value of y is skewed distribution.

dbcoffee
  • 173
  • 1
  • 8
  • Hi there, you will likely be able to find help here, but please try to post your code in a reproducible example with data: https://stackoverflow.com/help/minimal-reproducible-example – jpsmith Apr 28 '20 at 01:11

2 Answers2

0

If the data are positive or non-negative,y ou may want to consider other distributions, such as Gamma, inverse Gaussian, or Tweedie, available in SPSS GENLIN, which fits generalized linear models and generalized estimating equations (GEE) for correlated data.

David Nichols
  • 576
  • 3
  • 5
0

I am working on the non-normal distribution(which factor in skewness) for my PhD. We have done some work on skewed logit in health data, from which we have build on works of john Nagler, and Irvine Burr(famous for burrs 10 distribution). check out my preprint here https://www.researchgate.net/publication/342189268_Time_effects_of_bacterial_vaginosis_on_infant_morbidities_in_Kenya_assessed_using_modified_skewed_generalized_estimating_equations . since the non-normal are not incorporated mostly in statistical software, you can build your own functions in open source like R.

Sam Mwenda
  • 150
  • 8