Ancestry deliberately skews their results for all ethnicities. In any research study, you need the following: 

1) 400-1000 randomly-chosen samples in each group, including a control group where applicable (when I was in college, it was 1000); and 1:1 ratios across all groups (For example, if you have 400 Native American samples, you also need to try to have 400 Jewish samples, 400 Northwestern European samples, etc.).

2) Avoidances of confirmation and selection bias.

3) A confidence rate or interval of 95%.

4) Outliers noted as such instead of thrown out, which goes back to avoidances of confirmation and selection bias.

5) Full disclosure of potential conflicts of interest 

6) A hypothesis, a converse (or an inverse) hypothesis to what the original hypothesis is, and then a null hypothesis just in case the results fit neither the hypothesis nor the converse hypothesis. 

7) Other factors that ensure that your data is accurate, precise, reliable, valid, and extrapolate across general populations within a given group. For example, if you’re studying only British DNA, you need to make sure that your study is extrapolatable across Britain. If you’re studying DNA across countries, you need to make sure that your study is extrapolatable across the globe.