TO: Students in CJ 160
FROM: R. B. Taylor
DATE: 3/11/00
RE: Grading for Paper 1; Problem list for paper 1
The grading is based on five, equally counted parts of your answer:
1
2
3
4
references
where you could get up to a score of 1.0 on each part. To get your final grade I added these up, and multiplied by two. This gives you a score on a 0 to 10 scale.
But since this assignment turned out to be much more difficult than I had anticipated, these grades got curved. That curve appears below.
|
A grade on your paper of __ at least |
Corresponds to an adjusted grade of at least: |
|
6 out of 10 (D) |
|
|
7 out of 10 (C) |
|
|
8 out of 10 (B) |
|
|
9 out of 10 (A) |
I encourage you to review the regrading policy that is posted on the website. If you feel your paper was graded unfairly then you should talk to me and feel free to request regrading, following the steps laid down in that policy.
In your answers to the first paper assignment, many of you approached the questions in ways that were similar across different papers but not necessarily on-target. Rather than inflict even more of my horrendous writing on you all, I opted instead to just list the numbers of the problems that seemed to be relevant to your paper. The problems are listed below. The ordering has no significance.
The paper gives a general list of the limits of FARS data discussed by Jacobs, rather than focusing on the limitations shown specifically by the data shown in the last column of the data table. More specifically, the majority of drivers in fatal crashes do not provide BAC test results. If you add up the numbers starting with the row "test refused" down to the row "unknown" , and look at those as a percent of the total, it is 56%. In short, it was best to start the discussion by focusing on the data.
In reflecting on the gender differences, the paper presumes these data differences are driven generally by gender-based drinking differences, rather than drinking-and-driving differences. Gender-based drinking differences may not matter if driving behavior is not involved.
In discussing the differences by gender (or the total column), the paper overlooks that most of the time there are no data on BAC levels for one reason or another.
In explaining the gender differences in percent of drivers in fatal crashes testing positive, the paper argues that the difference is driven by the fact that there are more male drivers than female drivers, or more male drivers involved in fatal crashes. This argument, however, is irrelevant because the prevalence rate is a column percent, and thereby controls for varying group sizes. The column percentage gets at the percent or proportion of that group.
The paper treats the data as reflective of all drivers, rather than reflective of drivers "involved" in crashes where fatalities resulted.
The paper states that most of the drinking and driving accidents (should say fatal crashes) occur at a certain BAC level, while ignoring the large volume of missing data.
When the paper says that more men are involved in drinking and driving than woemn (and we really don't know this - we know that more men than women are drivers in fatal crashes), the paper overlooks the fact that more men than women are SAFE (BAC below .10) when they crash. In other words, comparing the numbers across is misleading; just compare the column percents (see 4 above).
The paper discusses data as though they were a sample; these represent all fatal crashes in 1998 in the participating states, adjusted up to capture the population. So strictly speaking, sampling error is not relevant.
Paper looks at just some of the missing data categories rather than all of them.
The paper treats count data as though they were rate data.
The paper presumes that we are learning about people (or the gender of people) who died; rather, the data are telling us about drivers involved in a fatal crash where someone (of unspecified gender) died or died later as a result.
On gender differences, the paper percentages across, rather than down. This is just telling us that there are more males in this category than females, for example, but it is NOT telling us about the proportion of males in this category.
For part 1 of the assignment, the writer fails to make specific reference to the numbers in the total column of the data table, and to link those numbers back to the relevant concerns about FARS data.
No calculation of column percentages for gender data (part 3).
Paper overlooks that the fars data are designed specifically to address accidents in which someone died. It is true that these do not tell us about nonfatal accidents, but if we focus just on fatal accidents, what are the limits?
On the references: * you needed to reference both the website for the FARS data and the Jacobs book; * both references needed to be complete. If you have questions about how to cite web sites check out the first link that says "how to cite them" under the section "evaluating web sources" on the "Criminal Justice" portion of my main website (http://www.rbtaylor.net/indexcj.htm)