A Student Guide to Research

This page gives a synopsis of strategies, techniques, and hints that will help you excel in your field of research as a student. The guide is generally written for graduate students, but the concepts presented here can be applied by researchers at any level, from high school to seniors in the field. My team and I have had success in producing research at top conferences using these methods, and I wanted to share this knowledge so that it will be useful for others. This is a living document, and I will update it as new resources or strategies become available. Each section contains a TL;DR summary.

Getting Started

Perhaps the most important and challenging task for beginning research is to pick a problem to solve. The best research problems are those that are 1) high impact, 2) clearly defined, and 3) of interest to you personally. Perhaps the worst way to start is to think about the technology or methodology without first clearly defining the problem itself. For example, if your thought process was to look at state-of-the-art methods in your field and then to come up with a new idea that improves upon the others, you might wind up searching for several years before stumbling upon a worthwhile topic.

Instead, it makes much more sense to list problems that are important or interesting to you and to later determine what methods or ideas could be used to solve those problems. If this "problem brainstorming" does not come naturally to you, don't worry. A great resource for finding problems is the "future work" or "limitations" section of papers in your field. I will cover literature review later, but find interesting papers, look at the future work sections, and compile a list of problems as outlined by other researchers if you find it difficult to come up with your own.

TL;DR

  • Start by defining the problem.

  • Pick something that is high impact, can be clearly stated, and is important/interesting to you personally.

  • If stuck, look at the future work sections of recent papers in your field.


Literature Review

Literature is incredibly important to ensure that your idea has not already been implemented by someone else. An essential resource for this is Google Scholar (scholar.google.com). Here, you can search for papers using a keyword search, and Google's indexing algorithm is incredibly good at turning up relevant results. Almost equally important is the "cited by" button, which appears as two small quotation marks underneath each result. This button returns papers that have cited the referenced paper. Moreover, checking the references/bibliography section of related papers can provide additional references.

The actual review strategy is also incredibly important. In general, I recommend reading papers in three steps: 1) abstract, 2) skimming, and 3) full-read. The abstract is designed to give you a high level overview of the paper, its contribution, and the results. This should give you enough information to understand whether the paper is relevant or similar to your own research goals. If the content only appears to be tangentially relevant, it's probably ok to skip further reading. If the paper does appear at least somewhat relevant, skimming is a good next step. To skim effectively, look at images (if present) and the "meat" sections of the paper first. These sections are usually titled methodology, infrastructure, design, experiments, or something to this extent, and they contain the most important information about the research content of the work. Look at each of these sections and skim them for keywords, paying attention to


TL;DR

  • Use Google Scholar or a reputable literature search engine

  • Read papers thoroughly only if they are relevant to your work. Skim otherwise.

  • Be thorough.


Ideation and Prototyping

Once you have picked a research problem, you need to figure out how to solve it. The approach you pick will depend highly on your field. Many fields such as biology are significantly based on experiments and testing, whereas fields like engineering often involve the development of new technology or algorithms with objective evaluation.

For the purpose of generalization, I will use the example of a bioengineering research problem that involves the development and application of a new technology to solve a problem in the vascular system. This example is fabricated (hypothetical), but will demonstrate the process consicely. Consider a group of patients that have a buildup of plaque in their arteries. You are a bioengineering PhD student and need to come up with a new way of addressing these plaques to improve blood flow and reduce heart disease. How would you approach the problem?

I suggest first taking time for ideation, where you write down and brainstorm a variety of different ideas or approaches. This process should be free of criticism or restrictions. Collect ideas even if they are somewhat ideal or not feasible with current technology. It is likely that your idea of a perfect solution will lead to other approaches that can be implemented practically. This approach of asking "what is the ideal?" will also give you a vision and a path to follow. If possible, try to draw from concepts or solutions that originate from other fields. For example: Would a nutritional approach be feasible? Would nanomachines that scrape the plaque off of aterial walls work? Perhaps certain bacteria could be genetically modified to consume plaque? Though some ideas might seem crazy, thinking outside the box can often lead to something that is actually feasible.

Once you have a list of ideas, including many off-the-wall or visionary concepts, then you can start thinking about feasibility. A good second step after ideation is the review and refinement of your ideas. Which ideas seem feasible? Is there a way to realize one of these strategies with technology from another field? Could some variations of these strategies work? At this point, you can start to rank your ideas by feasibility, impact, and/or intuition. The goal is to come up with a short list of just a few items that you can start to test or implement. Also consider the time it would take to test or build each solution when ranking your approaches.

At this point, you should also re-conduct a literature review to ensure you are trying something new!! Consulting with your advisor, mentor, colleagues, or collaborators is also recommended.

Finally, once you have an idea that seems feasible, rapid prototyping and review are key. I have seen many students that try to perfect an idea before they even start, only to find that they have spent too much time thinking about the idea. The process of implementation and prototyping is synonymous with learning. Trying to quickly build a prototype will accomplish several things at once: 1) You will quickly find out whether or not your idea is feasible. 2) You will improve your ability to quickly implement your ideas. 3) You will learn more about the problem itself, which can lead to new approaches. This process of prototyping, learning, and honing your skills is essential in research.

The key of this approach is to continue until you succeed. Continue coming up with ideas, trying to implement them, and learning from the experience, and you will eventually succeed. Persistence is key.



Testing and Refinement

under construction



Experimentation

Experimentation serves the purpose of providing evidence to support a claim or hypothesis. If you come up with a new method for interaction or a hypothesis for human behavior, you can test whether the interaction is effective or whether the behavior meets your hypothesis. Simple examples of experiments include:

  • Testing the effect of a new fertilizer on plant growth.

  • Evaluating a new algorithm to compare performance to a state-of-the-art technique.

  • Determining whether a vaccine is effective and safe to use.

  • Recording and comparing solar radiation data to determine if solar flare occurrences fit a particular distribution.

  • Obtaining data from students who are testing a new type of learning material.

There are many other types of experiments, but For simplicity I will provide detailed examples of systems tests, human subjects experiments, and clinical trials.

Systems tests typically describe a technical analysis of a new method in the fields of chemistry, engineering or computer science. For example, a student may come up with a new method for realistically rendering polygons on a graphics card, and he or she might claim that this method is more efficient than state-of-the art rendering techniques. To verify this claim, you could design an experiment that tests the performance of the method versus a current rendering technique from NVIDIA. A sample of this type of experiment can be found here.

Human subjects experiments are experiments that involve the use of human participants, which typically require an application to an institutional review board (IRB). IRBs are important to ensure that individuals are treated ethically and are exposed to minimal risk during experimentation. Next, you will typically recruit participants to test a method or hypothesis. For example, a researcher might want to test the usability of a new adaptive learning technique on student learning outcomes. To do so, 50 students from a college class could be invited to study for one hour a day over the course of six weeks, 25 of whom learn using the researcher's technique and 25 of whom learn using a common-practice or best-known technique. The researcher could administer a series of quizzes or tests over the course of the study to evaluate retention or understanding, which would provide evidence as to whether the researcher's technique outperformed the other technique. Data analysis is covered in the next section.

Clinical trials are a type of human subjects experiment that often involve the testing of a medical treatment to determine its effectiveness. For example, many researchers and clinicians conducted clinical trials to test the effectiveness and safety of a vaccine for the COVID-19 virus. These tests recruited volunteers that were willing to receive the vaccine, and the researchers then measured the infection rates of individuals in a vaccine group and a placebo group. In comparison with other human subjects experiments, clinical trials are often much more rigorously monitored or reviewed because of the possibility of negative outcomes such as side-effects and often because patients may depend on the treatment for their well-being.

Analysis

Analysis, in particular statistical analysis, is a fundamental part of testing a given method or determining whether a variable influences an outcome. For example, in the example of testing the effectiveness of a vaccine (as described in the experimentation section above), researchers need to test whether the vaccine 1) is safe, and 2) is effective at preventing infections.

To test this, the researchers might have recruited a group of participants that were divided into two groups, a vaccine group and placebo group. This means that one group (let's say 1000 individuals) received the vaccine, and the other group (also 1000 individuals) received a placebo such as sugar water or saline. The researcher's hypothesis is that the vaccine will be safer and more effective in preventing infection than the placebo.

After a year, the researchers obtain data from all the participants to determine 1) the incidence of side effects and 2) the incidence of infections for both groups. In the vaccine group, 12 infections occurred, and 3 individuals claimed to experience side effects. In the placebo group, 167 infections occurred, and 4 individuals claimed to experience side effects. At first glance, it may seem that the placebo group had greater than 10 times the number of infections as the vaccine group. Although this is true for the individuals that took part in the experiment, we need to go a step further to determine if these occurrences happened by chance or if the results are actually representative of the general population. For this purpose, ANOVA, or analysis of variance, is incredibly useful. Using ANOVA, we can conduct a test, such as an f-test, that determines the probability of whether the differences happened by chance. The type of test used can vary greatly with the type of data or experiment design you choose, so be sure to search for the appropriate test for your experiment design.

Writing

Effective writing is perhaps just as important as conducting research itself. Clear and well-organized papers are much more likely to get accepted and will have a greater impact on the research community through their ability to disseminate information.

In order to choose an effective writing style, I recommend first looking up highly cited papers or those that have won awards in your field. This will give you an idea of the structure, wording, and format used to present research information effectively. Though you should not copy text word-for-word, it would be a good idea to emulate the structure of one of these papers that best matches your own writing style.

In addition, writing should occur not after the research is complete, but before the research even starts. Even during the literature or brainstorming phases of your research, you can start writing sections like the abstract, prior work, or introduction that are often found in research papers. Writing early will not only help you focus your ideas, but it will prevent you from having to do all of the writing in the last minute. Scheduling 1-2 hour blocks of time every week for writing alone is very useful for this purpose. I recommend doing this just after you finish research every week so that content is still fresh on your mind and easy to put into written form.

Submission

under construction



Ethics

under construction

Examples of what not to do:

  • Salami submission

  • Double submission

  • Faking results / false claims

  • Exaggerating

  • Inappropriate use of funds

  • Gift authorship



Will my paper be accepted?

A clear answer to this question does not really exist. The review process for conferences differs greatly between fields and disciplines, and the sheer quantity of papers submitted to conferences is making acceptance increasingly competitive. In all likelihood, some of your papers will probably be rejected. However, there are several things you can do to improve your chances.

A great way to understand the standards for any conference is to look up the best papers for that conference for the past several years. If your paper exhibits similar qualities in terms of novelty, depth of research, and technical correctness, your chances are probably good. Secondly, look up the reviewer guidelines for the conference to see how your paper will be judged. It even makes sense to review your own paper using these guidelines and to write out why your paper would be a clear accept. If this is difficult, you may need to re-evaluate what you are submitting. Finally, reach out to a colleague to give your paper a high level review. He or she should ideally be a senior researcher in your field and should be outside of your organization. Not only will this result in some initial objective comments, you should get an idea of how your paper might be rated from someone else's perspective. Lastly, ensure that you really have done a proper literature review. Many good papers are rejected due to the fact that they overlooked prior work that achieved similar results or developed a similar method.

TL;DR

  • Look up best papers for the conference to which you plan to submit and compare to yours

  • Review your own paper using the guidelines published on the conference/journal page

  • Ask a colleague in your field (preferable a senior researcher outside of your organization) for feedback