单因素方差检验

一个比较初级的统计问题：
新药研究，80名高血脂患者，随机分成4组，进行双盲实验。6周后测量低密度脂蛋白。研究4组的指标有无差异。
1.各组样本不到30个，是否要进行正态性检验？
2.4个组要不要进行方差齐性检验？

The sample size, statistical analysis, interpretation, and assumption as you enquired are highly driven by your objective and hypothesis.

For lipid lowering clinical trial, the classic endpoint is percentage change from baseline on low density lipoprotein cholesterol level. If you are using this specific outcome, normality check can be very appreciated. However, you need to be very careful that once you fail the normality assumption, the conclusion drawn from invalidate model is questionable.

So, I suggest you pre-specify what kind of analysis you will perform first. And followed by normality check, if succeed, that's the end. Otherwise, you can perform sensitivity analysis such as robust methods to support your argument.

Outside the business here. Don't know why you have 4 group rather than 2, pay extra attention on the multiplicity problem. And do not expect you can observe meaningful adverse experience. Let's face it. 20 per group may be statistically powered, but not clinical significant at all.

Have a good day.