GWAS has been done with family-based samples (e.g., data from the Collaborative Study on the Genetics of Alcoholism [COGA] study). What are the strengths and disadvantage of conducting GWAS that includes related individuals?
in practice very few studies recruit cases and controls any more and we use everyone available
I think the “best” in terms of power is 1:1 - it maximizes the variance of the binomial. There may be reasons to deviate, e.g., if it’s easier to get controls
All the things that happens in the real world GWAS 
many cohorts use relatives - there are no disadvantages as long as you use software designed for this
there can be advantages but most software is designed to give a standard gwas and doesn’t provide the advantages
To be clear you CAN get 2 types of estimates - most GWAS with relatives do not plan to do this and want to estimate beta and se that is equivalent to an unrelated sample
I think the GWAS has more statistical power if including more samples like related individuals
that’s right! Usually, unless there are lots of relatives, relatedness is a nuisance. Not accounting for it will lead to an anti-conservative bias in the SEs, but it won’t necessarily bias the point estimate of the beta
Anybody who has missing data on the phenotype or on a covariate that you use in your model will be excluded from the GWAS by Plink.
family-based samples allow “within-family” estimates of GWAS betas. Those are protected from many of the biggest confounders in population-based GWASs (popstrat, genetic nurture, assortative mating). But they tend to be less powered - effective N is about 1/6 to 1/3 depending on the within-family design
actually the effective N is usually slightly less than the same N of unrelated people - but the difference is usually small