How do I generate an Undergraduate Thesis worthy of publication?

Question

Simply, what do you expect from a good bachelor's thesis?

Answer 1

What do you expect from a good bachelor's thesis?

The expectations for a bachelor's thesis vary from country to country and from field to field. However, I think the main points can be summarised as follows:

• the author shows a good understanding of the given problem
• the author is aware of the existing literature in the field, can discuss the literature critically and understands where their problem sits in the literature
• the thesis is well-written and successfully communicates the author's findings, ensuring that anyone wishing to replicate their experiments could do so

A bachelors thesis is NOT expected to be an entirely original work nor does it need to represent an advancement in the field. I think I am safe in stating that publishing work from a bachelor's thesis is highly unusual.

A successful thesis requires consistent work on a specific problem with demonstrable results that can be disseminated to the community (be this your peers at university, future employers or maybe to a journal for peer review).

You are missing one key ingredient: a supervisor.

At the bachelor's level, it is very difficult (due to inexperience or unfamiliarity with the literature) to identify a problem which is open-ended enough to allow plenty of work on it, but not so esoteric or challenging that the student will not be able to make any meaningful progress or learn anything valuable from the experience.

You state that

Most of my lecturers are PhD scholars themselves struggling with their own papers. Ego in academia may prevent them from assisting me in my paper.

Have you asked them about this or are you making an assumption? They may well be very happy to supervise your project; after all, getting other people interested in research is very important for academics.

I think that you and your thesis will be far more successful if you find a supervisor who can give you the guidance and direction you need. Your supervisor will also be able to answer your questions about publication of your work (if it merits it-- again, I stress that publication should be low on your list of priorities. Get to know the literature, gain some expertise surrounding your problem and write, write, write).

Now to address your questions about an Ig Nobel.

I'm not sure if you've maybe misunderstood what these prizes are about: they're a gentle satirical mocking of the Nobel prizes, and generally awarded for ``pointless'' or silly research-- not serious science. I'm not sure that aiming to be awarded one for your bachelor's thesis is a good idea.

You can nominate yourself for an Ig Nobel, but as their own website states (emphasis mine):

Self-nominees seldom win. It seems characteristic of the Ig Nobel prizes that the "Igginess" is a side-effect, not a goal.

Don't aim for an Ig Nobel. Concentrate on finding a supervisor, working on a tough problem, writing a great thesis and getting to know the joys and pitfalls of research.

There are some other questions and answers on this site that may help you further:

1. Finding a supervisor (this question is about getting a supervisor at another university, but the advice is applicable to your situation).
2. Do's and don'ts of undergrad research
3. What counts as undergrad research (when aiming for grad school)

Answer 2

I'm trying for an Ig Nobel.

First, let me say that I love your attitude. Academia could stand to enjoy a bit more humor, especially when applied productively. Some of that's been lost when people are afraid to step out of place or appear unprofessional; but, if you're aiming for the stars, you'll be in good company if you're relaxed.

And since

I'm currently a CompSci&Engg

, I'd recommend:

Do p-hacking!

Say that you want to find some desired relationship in your data. You do a regression, but the results aren't really convincing. What then? Well, clearly, you just picked the wrong regression model. So, you try a bunch of 'em 'til it works, including a bunch of correction methods for confounders, and once you find a statistically significant result, clearly that was the correct model all along. p<0.05, because it's science, yo.

Clearly I'm joking, right? Unfortunately not. There're a lot of unbound degrees of freedom in practiced experimental design and data analysis; these unbound degrees of freedom are optimized for reportability, leading to skewed data analyses. It's like doing machine learning, except everything's training data; no validation set until someone comes along to do it later.

It seems perfect because:

1. It's an extremely important topic.

2. It's under-researched.

3. It's funny.

1. It's important.

From "Experts issue warning on problems with P values" (2016):

“It is a safe bet that people have suffered or died because scientists (and editors, regulators, journalists and others) have used significance tests to interpret results,” Rothman wrote.

They're not being overly dramatic. A lot of medical literature reports findings that just don't follow, guiding future research in fruitless directions as people who have medical problems today spend their time downing snake oil.

For a quick intro, the American Statistical Association (ASA) provides "The ASA's Statement on p-Values: Context, Process, and Purpose" (2016). The overall problem's called the reproducibility crisis.

2. It's under-researched.

While we know that it's a huge problem, I haven't seen good studies into how it keeps happening or effective initiatives to improve the situation.

For an undergrad thesis worthy of an Ig Nobel prize, you could do something like come up with a clear, consistent methodology for turning arbitrary experimental data into results that'd be considered valid by currently practiced standards. Ultimately, your goal is to produce a methodology for "proving" any conclusion you like with p<0.01, or something like that, then use this methodology to "prove" all sorts of absurd results.

Off-hand, I'd try this:

1. Get an experimental data set.

2. Try to fit it to a regression model.

3. If the results are "significant", record.

4. Repeat Steps (2)-(3) once for each possible combination of:

   • Regression model. (Use linear, exponential, log, polynomials, etc..)

   • Correction method.

   • Outlier reduction.

   • Variable transform.

5. Repeat Step (4) for various experimental setups.

6. Report the most "statistically significant" result.

7. Include disclaimers about how the results are preliminary and that more research is needed to determine significance.

3. It's funny.

If it's on xkcd, it's funny. (Proof: Randomly sampled a few comics; analyzed for various definitions of "funny" 'til I found one that fit; p<0.05.)

Seriously though, this is the sort of project that if you do well, then add a bit of flavor with funny comics that anyone can understand (like this xkcd), you could easily make headline news. Plus a few new memes!

4. ???

Do it for the memes, man. If you don't, who will?

5. Profit: Ig Nobel.

If you write a thesis on how to "scientifically" prove anything, then use it to prove a bunch of silly non-sense, it seems like you should get an Ig Nobel. Plus a lot more fame; again, it's an extremely important topic that needs more public awareness, both within academia and outside of it.

If they don't give you one, I'd write you up an Ig Ig Nobel. I'll even print it off on the back of a napkin, egregia cum laude, if the thesis is good enough!