search

easystats / parameters

:bar_chart: Computation and processing of models' parameters
https://easystats.github.io/parameters/
GNU General Public License v3.0
440 stars 36 forks source link

DF between-within #192

Open strengejacke opened 4 years ago

strengejacke commented 4 years ago

http://de.saswiki.org/images/d/d4/11.KSFE-2007-Buescher-Analyse_longitudinaler_Daten.pdf

mattansb commented 4 years ago

Ich nicht spreche Deutsch 🤷‍♂️ Was ist das? (Hope I got that right?)

strengejacke commented 4 years ago

I haven't found a good resource quickly, the idea came up when reading this thread: https://stats.stackexchange.com/q/443241/54740

strengejacke commented 4 years ago

http://www2.sas.com/proceedings/sugi26/p262-26.pdf

Generalized commented 4 years ago

Also, you may find this one informative: Li, Peng & Redden, David. (2015). Comparing denominator degrees of freedom approximations for the generalized linear mixed model in analyzing binary outcome in small sample cluster-randomized trials. BMC medical research methodology. 15. 38. 10.1186/s12874-015-0026-x.

https://www.researchgate.net/publication/275363360_Comparing_denominator_degrees_of_freedom_approximations_for_the_generalized_linear_mixed_model_in_analyzing_binary_outcome_in_small_sample_cluster-randomized_trials/stats

Currently, in R, only the containment method seems to be implemented (glmmPQL, nlme). In the repeated observations, especially in longitudinal trials, the between-within DF is often preferred as the last-resort option, if K-R or Satterthwaite cannot be used.

strengejacke commented 4 years ago

Thanks for the hint to the paper. I'm not sure how the DDF-B-W are calculated. The paper simply states K - rank[x], but I'm not sure where rank[x] refers to.

strengejacke commented 4 years ago

Any SAS-Users here that can provide the output of some toy example that can be reproduced in R, so I can check the implementation of dof_betwithin()? I think in particular for nested random effects, the computation might differ...

strengejacke commented 3 years ago

@bwiernik maybe you could also give some input here?

bwiernik commented 3 years ago

rank(x) refers to rank of the fixed effects design matrix

bwiernik commented 3 years ago

The between-within method was derived for repeated measures (fixed effects) designs with various error structures, not for random effects or mixed effects designs. So it doesn't pay any attention to the random effects design matrix