We tested the associations for effect modification by sex and race through the addition of an interaction term between each predictor and either sex or race in the model, using an α cut-off of 0.05.
There was a significant statistical interaction involving sex and race in the model for thyroid cancer.
2-way interactions between physical activity and age, gender and race in the model adjusted for percent fat were tested.
Then when adding race to the model (model 4), the HR for the lowest SES group was non-significant (HR = 1.03, 95% CI: 0.97-1.08).
I have not studied whether conventions produce bounces in Congressional races and the model makes no assumptions about it, but it is a reasonable hypothesis.
By the 1960s fashion shoots meant 35mm action cameras chasing leaping girls across flowery meadows or race tracks: the models even had expressions on their faces.
Adjusting for race in the models showed that Malays had a higher OR than Chinese.
We also dropped "race" from the models due to the existing heterogeneous nature and insignificant sample size of the non-Caucasian group.
An additional concern was that inclusion of race in the models, given its correlation with SES, could artificially inflate the exposure RRs for the predominantly white exposed population.
We also conducted analyses by including age, sex, and race in the models and found that the associations did not change substantially (data not shown).
