Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Scott Onestak

Pages: [1] 2 3
Paper Replication and Project / Re: Improved Replication
« on: May 04, 2017, 08:21:24 PM »

So, it's really hard to help with getting your replication closer to the researcher's values because we don't know what you have already done.  Also, nobody has the perfect solution.

What I would say is: if your replication was close enough to receive near full marks on the assignment, then it should be good enough to use on the critique.  Even if one or two of the coefficients is off, just ask yourself if that variable is an important variable in your critique.  If it isn't the subject of your paper, then I don't think it is that big of a deal.

Paper Replication and Project / Re: Sources
« on: May 04, 2017, 08:13:26 PM »

It's difficult to know how many sources a paper should have because all sources are not created equal.  Some sources have a vast amount of relevant information while other sources may only contain a few lines of material.

Too few sources, and you won't have enough content.  Too many sources, and your paper may be redundant.

Find a good middle ground where you're presenting a complete argument without repeating yourself, and only you can really know how many sources you need for that because everyone is working on different topics.

Paper Replication and Project / Re: Question about paraphrasing
« on: May 04, 2017, 08:01:41 PM »

So here's a basic rule of citations: if it's your own idea, you don't need a citation.  If you take any idea that is not fully your own, then you need to cite it.

Everyday, people have the same ideas, and it's not plagiarism.  However, if you're looking at something and then coming up with an answer that is essentially the exact same thing, it would most likely be wise to cite your source.


To be frank, if you cannot get results, then the paper will have to be a theoretical paper or explain why the results you got are the way they are.  I think you're asking more about "what happens if I cannot get strong, conclusive results?"

Based on everything you have learned this semester, you should be able to determine if something is statistically and/or economically significant.  With the results you have, your paper should be able to explain why there is omitted variable bias and what it is, what your proxy is and how it solves that problem, and why the results show some significance.

When using a proxy variable, even if something isn't a statistically significant deviation, you can still make your argument that there is omitted variable bias by stating why you think the proxy is only picking up some of the total omitted variable bias effect.  So, there are many options for you to go with to achieve promising results.  Your objective should be to defend the strongest argument you have.  Sound, conclusive data-driven statistics often make for the best arguments because if the analysis is done correctly, it is hard to dispute them.  However, a well-delivered theoretical argument can be just as effective.

Paper Replication and Project / Re: Wrong Year
« on: April 29, 2017, 11:46:53 PM »

This is really a judgement call, and without any further information on the variable, nobody but you is going to have any way of determining the best solution.  The question I think you should ask is: does the year make a difference in the variable value?

For a variable like race that doesn't change over time, it's pretty easy to justify the inclusion of the variable despite the year it was taken.
If the variable could change values between 1979 and 1980, then you may have a harder time justifying its inclusion.  If you do include the variable, you should justify its inclusion in your paper.

Paper Replication and Project / Re: Difficulties with getting results
« on: April 29, 2017, 11:40:46 PM »

My suggestion for learning more about the coefficients in the code book would be to use the link provided in the replication tips to find more detailed explanations of the variables you are looking at in the online code book.

I would caution that you seem to really be pushing the boundaries of good research.  While it's great that you've though about this omitted variable bias thoroughly to come up with a solid hypothesis, you cannot let your desire for justifiable results cloud your judgement of good, unbiased research.  If there is a reason that one proxy should be used over the others, then use it.  However, you shouldn't twist your research and paper into a pretzel to justify your hypothesis.  You'd be better off writing a really solid theoretical paper on the omitted bias than presenting questionable research to the grader. 

Paper Replication and Project / Re: Question on citations
« on: April 28, 2017, 12:50:13 PM »
Yes, it is not your original idea, so you should cite it.  Remember, you only have to cite stuff like this (not a direct quote), before you move onto another source.  So, if all of this stuff is coming from the same page of the research paper, you would only need one citation if all those sentences are grouped together in your paper.  Just remember that information found on different pages of the research paper would require a different citation and a new citation is needed if you start another paragraph.

Paper Replication and Project / Re: Question on citations
« on: April 27, 2017, 09:03:15 PM »

I'm not positive what citation style you're using, but it sounds like Chicago to me.  No matter the citation style, you should always cite your source when you use a direct quote or are paraphrasing.  If just referencing, like "these results are consistent with the results found in this paper," I don't think a citation is necessary.  However, pulling any information or idea that is not your own should always be cited.

Here's a quote from the Chicago citation manual:

" should include a note (endnote or footnote) each time you use a source, whether through a direct quote or through a paraphrase or summary. Footnotes will be added at the end of the page on which the source is referenced, and endnotes will be compiled at the end of each chapter or at the end of the entire document."
[/size]If you want to read more, this is a good resource: [/size]

Exams and Grading / Re: Material Qj
« on: April 12, 2017, 03:18:20 AM »

So, I can't tell you exactly how to do part j, but I'll try to lead you in the right direction.  Part j is combining multiple statistical concepts you've learned, so break it down into its parts. 

1) What would you do in order to conclude that each extra unit of x1 accounts for an increase in y of at least c? 

2) Obviously, at least one of these components has to be calculated using the number of observations, n.  Otherwise, the question would be impossible.  Therefore, with the numbers and information provided to you on the exam, how could you change n so that x1 accounts for an increase in y of at least c units.

For your second question, this gets at the heart of the question because to answer yes or no would tell you how to construct your model.  Having completed the replication and having learned what you have about constructing linear regression models, how would you determine the causal effect on wages (percentage variation) of working in different occupation sectors for women with men now in the data set?  What changes would have to be made to the original model, if any, to accommodate for this new factor?

Material / Re: Paper Question Part 2 question f)
« on: April 12, 2017, 03:02:09 AM »

I'm not completely understanding your comprehension of this question, but this might help you understand what part f is getting at.

You have your model that's been constructed.  What test would you be able to construct that can tell you that there is no effect on wages for obese women employed in either sales or services (relative to the production sector).  Beyond that, it's hard to help without giving away the answer.  But this is the key phrase you should be focusing on.

Exams and Grading / Re: Question 1f and 1g
« on: April 10, 2017, 10:11:35 PM »

I suggest looking over notes 10, section 5 for more information about interaction terms.

Remember that interactions are interpreted as premiums, so when thinking about whether to incorporate an interaction or not, a good starting question might be to ask if you expect there to be some premium for individuals in the data that is not simply captured by the variables already.

For interpreting and the significance, the notes do a firm job of securing that foundation.

Exams and Grading / Re: Paper question j
« on: April 10, 2017, 08:47:04 PM »

Without knowing what the variable it is, it is hard to say whether a variable is related to the treatment or superfluous.

Here's an example:  The paper we used for my class was wage discrimination between men and women two years out from college.  A similar scenario to what you are asking would be if some variable for the employer can be associated with the treatment variable for woman.  In labor economics, there's a tactic used by some employers to motivate and keep workers in the firm, known as sequencing of pay (very much used in the law profession).  If firms use this, employees are often underpaid for their skills at the start of their careers to be overpaid later on.  If we had a variable quantifying the sequencing of pay or if a firm used it, then it may be possible to associate this variable with the woman variable.  It may be more likely for women to go into fields that have sequencing of pay or it may be less likely.

Therefore, even though a variable may not seem to be clearly associated with the treatment or outcome variable, it may very well be.  Your responsibility for this problem is to determine whether that variable is relevant and should be included in the model.

Material / Re: 2 questions about paper question
« on: April 10, 2017, 08:19:59 PM »
Hey Shenxiong,

Nobody is going to be able to answer your question without directly giving away whether you answer is correct or incorrect.  Therefore, we cannot really answer the question any further besides pointing you in the direction of the notes, which Alexis has already done.

Material / Re: Paper question part (i)
« on: April 10, 2017, 08:15:23 PM »

Yes, you are thinking about selection bias correctly.

To help, I would say to think about who is most likely to complete and return a questionnaire.  Beyond that, I cannot really discuss the problem any further without giving away the answer. [/size][size=78%] [/size]

Material / Re: Questions about paper questions
« on: April 10, 2017, 08:09:28 PM »

I would say: no, a) a) and f) are not the same thing, though they may be very similarly related.  Question a) a) is asking for a model while f) is asking for a test for the hypothesis given.

Similar to what Thomas said before, besides providing that distinction, I cannot really discuss much more about the questions without giving away the answers.

Pages: [1] 2 3