1. True/False/Uncertain - Brief explain.

3.If one uses US states data to model COVID-19 cases, one should use an offset.

I'm not sure about my answer. I answered this.

True, In the US, it has 52 states and each state has a different population. For instance, NYC has around 8 million people, and LA has around 3 million people. So, if we compare the number of Covid cases, of course NYC has more Covid cases than LA because of the population. In this case, we need to use offset because it is not comparable. If they have the same population, then we don't need to offset it.

4. Global homogeneity SSG is never zero when we start with each observation in its own group with the Ward clustering method.

True. In the hierarchical n groups to start each one of the observations has their own group. We start with 1 person in their individual group then we aim to group them under one single group. Hierarchical method using WARD method, SSK is the group homogeneity, if SS1 has only 1 observation with the group is going to be "0" at the beginning. Starting with homogeneity of "0" all of these different sums of squares are 0 because I have 1 observation.

5. Having more missing values in the variable education for higher levels of education is an example of MNAR.

True / uncertain. Missing not at random (MNAR), depends on missing values. Some people will talk about their higher education. Those who refuse to provide this information because they did not have high education. Similar examples with high income, people who have high income tend to not answer about this sensitive information, so that's why there are lots of missing values. In this case, people who archive minimal for their education tend to have lower rates of responses.

Thank you