Approve web version of data cleaning lab

[rpruim]

As a prototype, I'm going to edit the data cleaning lab

[rpruim]

Some notes as I went through the student worksheet part:

• [x] "bad" data? Is that a good term?
• [ ] Not sure it belongs here, but the distinction between clean and tidy is important.
• [x] field is not a commonly used term in statistics. The usual term is variable. I think field is primarily used in the context of some databases (but I'd like to be reinforced on that feeling). This is a small thing, but it could be a distractor for students/faculty. There will probably be other situations like this as we go along. I don't have a good solution at the moment. But I decided to give each variable/field a formal name (like you would do in statistical software or a computer program). This reduces the number of times variable/field needs to be used -- and it also make things look more like actual data.
• [x] the data set description looks like the data set would include only students who leave a university -- would it be better to have a data set that includes both students who do and do not leave? I've added a variable for withdrawl -- it is easy to remove again.
• [x] I get nervous about casual uses of the term "outlier", especially without more context.
• [x] Not all incorrect data values are outliers (at least in the most naive sense of being especially large or small) -- should there be a discussion of this problem? Also, if we investigate "outliers" more carefully than other values, then we are more likely to detect some types of erroneous data than other types.
• [x] The "ethical issues" section seems tilted toward CS in a few places. It seems targeted toward creating software (products, users) rather than doing data analysis (analyses, clients, subjects). Since the point here is about data cleaning (and subsequent analysis), these could be edited to make them more neutral or to lean into statistics/data science.
• [x] The anonymity question is focussed on researchers, but in many cases it is required that the researchers have access to identifiable data to do their work. This question could be expanded a bit. Data access and security, IRB, consent, etc. could all be included.

Resources?

• This blog post talks about some articles that try to estimate how many people in the US are uniquely determined by zip, sex, and birthdate: https://godplaysdice.blogspot.com/2009/12/uniquely-identifying-people-by-birth.html

[LoriCarter]

Yep! Written by a computer scientist! Feel free to make it correct for a data scientist

[rpruim]

Notes from Instructors page

• [x] I think 'replicate' is more commonly used than 'duplicate' in this context. Does someone have evidence to the contrary? (I know it is that way in the circles I'm in.)
• [x] Regarding outliers: It might be good to add some more discussion about the use of the word "outlier". It seems to be used here to mean "extreme value". Every distribution has smallest and largest values, that does not make them outliers, they may very well fit the pattern of the data -- especially if we consider marginal distributions. For skewed or long-tailed distributions, there may be appropriate analysis methods that can handle data that don't look "normal". In general, beginners are perhaps too eager to throw things out.

[rpruim]

@LoriCarter, I'll do a little tweaking of the issues list. But we may not want to lose the current list. Does it appear in some other labs? [I could imagine this list appearing in multiple labs, perhaps with slightly different versions depending on the focus of the lab or the intended students.]

[rpruim]

I've edited the issues list. I'll push shortly. Here are some comments.

Here is the previous version for reliability:

Reliability: Ensuring that one's work is consistently functional as specified through adequate testing/duplication with a variety of users.

As written, it doesn't work so well for data analysis. Ideally I'd like something that works equally well for computer and data sciences. Does this work?

Reliability: Ensuring that one's work consistently does what claims to do.

It doesn't say as much about the how, but I think the goal works for multiple disciplines.

I'm also wondering how to better distinguish between Algorithmic bias (which perhaps is just plain bias) and Data integrity, which also mentions bias. I have some ideas here and I'll see if I can draft something reasonable.

OK. Here's my current draft:

Bias: Recognizing and reducing any potential biases present in an algorithm, model, or process that may have arisen as a result of ignorance, assumptions, or past discriminatory societal patterns.

Data Integrity: Collecting and handling data in a way that accurately reflects the phenomena being studied and is appropriate for the analysis techniques employed.

Can someone help me distinguish between professional citizenship and professional ethics?

[rpruim]

You can find the full issues list at https://wonderful-fermi-8c4db8.netlify.app/labs/data-cleaning/data-cleaning-worksheet/ or in the corresponding place in the repo.

@kcarnold, we should chat about

• the best way to handle multiple versions of a lab sheet. I have placeholder links for Word, PDF, and Rmd, but I haven't investigated how best to deal with this in blogdown.
• formatting of the HTML lab sheets. I'm guessing we want a different template document (we don't need to have the date displayed, for example, on the student handouts -- or at least not so prominently).
• naming conventions. The URL above is a bit redundant. data-cleaning/worksheet/ would be nicer, but then we have tons of tiles named worksheet.*. It would be nice if the output file names/URLs didn't have to match the input file names. Do you happen to know if blogdown supports that?

[ccrocke1]

Hi! Quick answer to the last question: professional citizenship: is contributing voice and opinion within one's expertise regarding the ethicality of the work of those around them. (dramatic example: Sally Clark's case in UK).

Professional ethics: contributing one's best effort in the workplace while respecting one's peers and upholding the requirements of one's employer and the discipline's overarching guild.

Hope that helps! Catherine

On Tue, Jul 6, 2021 at 2:18 PM Randall Pruim @.***> wrote:

I've edited the issues list. I'll push shortly. Here are some comments.

Here is the previous version for reliability:

Reliability: Ensuring that one's work is consistently functional as specified through adequate testing/duplication with a variety of users.

As written, it doesn't work so well for data analysis. Ideally I'd like something that works equally well for computer and data sciences. Does this work?

Reliability: Ensuring that one's work consistently does what claims to do.

It doesn't say as much about the how, but I think the goal works for multiple disciplines.

I'm also wondering how to better distinguish between Algorithmic bias (which perhaps is just plain bias) and Data integrity, which also mentions bias. I have some ideas here and I'll see if I can draft something reasonable.

OK. Here's my current draft:

Bias: Recognizing and reducing any potential biases present in an algorithm, model, or process that may have arisen as a result of ignorance, assumptions, or past discriminatory societal patterns.

Data Integrity: Collecting and handling data in a way that accurately reflects the phenomena being studied and is appropriate for the analysis techniques employed.

Can someone help me distinguish between professional citizenship and professional ethics?

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/integrated-ethics/labs/issues/24#issuecomment-875088628, or unsubscribe https://github.com/notifications/unsubscribe-auth/AUU472CQT5XQ6GRFW6OB4TDTWNXLXANCNFSM47452TZQ .

-- Catherine Crockett, Ph.D. Professor of Mathematics Mathematical, Information and Computer Sciences Point Loma Nazarene University 3900 Lomaland Drive San Diego, Ca 92106 619.849.2723 @.***

[rpruim]

Are these canonical terms? I find them confusing since citizenship seems like it should be about more than just "contributing voice and opinion... regarding ethicality". In fact, if the terms and definitions were swapped, it would make as much or more sense to me.

The mathematics side of me says it doesn't matter what words we use as long as we define them. The pedagogy side of me thinks that the names we choose also matter.

I wonder if we could at least avoid using "ethics" in the terminology. Is the second items just plain "professionalism?"

[rpruim]

Just to be clear -- I did see the definitions. I'm just trying to figure out (a) exactly what distinction is being made and (b) whether we have the right names for the distinction.

[rpruim]

I did another round of edits. I think all of my check boxes have now been addressed at least somewhat. Some of the broad strokes:

• More care about the term outlier.
• Much more caution about "filling in" values and a brief mention of imputation methods in the instructor notes.
• Less focus on editing bad data, more focus on consequences.

This could perhaps be improved a bit more by making a tighter connection to the ethical issues, but I'm ready to close this now. @LoriCarter, you might want to take a look when you get back to make sure you are still happy with my changes.