Please write one separate response to each of the 2 posts below. I attached the 2 questions that the posts below are answering.
Post #1 (Wesly):
There are a few biases that need to be addressed in the proposed study. First off the researcher didn’t collect education level. Educational level can be a big influence on a pay scale. Secondly the researcher didn’t ask for experience level. Someone with 10 to 15 years of experience has the potential to earn more than someone just out of college or 2 to 3 years of experience. Lastly the researcher didn’t ask about years with the company. In my experience despite overall experience, when you join a company it is usually close to the bottom and you must work your way up. With the information that the researcher collected I don’t believe a firm conclusion can be made, however if you add a regression indicator variable in the equation for these other factors, a strong conclusion can be made.
The researcher has collected more information than he needs to determine the results. The researcher should see if there is any data that he can omit. To be able to omit certain variables he must for ask if any of the data meet the two conditions: (1) is the regressor, i.e. X, correlated with the omitted variable, and (2) is the omitted variable a determinant of the dependent variable, Y (Stock & Watson, p. 182). If these two conditions are two the researcher can omit the data that will not affect his results.
Stock, James H., & Watson, Mark W. (2011). Introduction to Econometrics, Third Edition. Boston, MA: Addison-Wesley
Post #2 (Pedro):
Based off of the given information, the researcher may fall victim to omitted variable bias. Our textbook, Introduction to Econometrics, states that omitted variable bias occurs when the regressor is correlated with a variable that has been omitted from the analysis (Stock, 2011, p. 180). To put this in clearer words, it is possible that the gender of the engineers is highly correlated with an aspect that they are not looking at, like education level. This would mean that maybe the females who work in the engineering department have less education than the males in the engineering department. This would result in the women making less than the men, but it wouldn’t be a result of gender bias in setting wages. Other than education level, it would be worth looking at other factors as well, such as time with the company, or performance review scores. Additionally, if the researcher truly is concerned about gender bias, then they should be looking at all departments, and not just the engineering department.
Stock, J., & Watson, M. (2011). Introduction to econometrics (3rd ed., p. 180). Boston: Addison-Wesley.
The researcher in this scenario has enough information to determine accurately whether time spent in prison has a permanent effect on a person’s wage rate. However, they need to be careful to ensure that they do not have too many variables. This matters because it is possible that multicollinearity exists between multiple variable, or regressors. This occurs when one regressor is highly, or perfectly correlated to another regressor causing a linear combination with each other (Stock, 2011, p. 199). Based off of the data the researcher is collecting, I don’t think any of them are too highly correlated with each other to cause a problem. It will just be important to to look at like for like data; as in looking at each variable while holding other variables constant.
Stock, J., & Watson, M. (2011). Introduction to econometrics (3rd ed., p. 199). Boston: Addison-Wesley.