Chapter 5

[YinghuiZhouu]

Aims:

• [x] 5.1 showing bias in small samples. Find the right sample.
• [x] 5.2 looks fine. How does the bias change when the compliance rate change
• [x] 5.3 Important. excludability violation.
• M1, M2, M3. M3 the only difference: U2. Could be a single parameter. Make it simpler.
• [x] 5.4 $cov(\hat{CACE}, \hat{ITT_{D}})$. combine with 5.1_sample_size. Point should be generated.

[YinghuiZhouu]

• [x] 5.2 add uncertainty and use direct labels. Change the sample size to 800 to show CACE becomes unbiased much faster.
• [x] 5.3 vary sample size instead of the compliance rate. Only CACE.
• [x] 5.4 add uncertainty
• [x] 5.5 add uncertainty and use direct labels. Change the sample size to 800.