How to research

[FailureToThrive]

I'll be working on a research project this summer. It will be my first time analyzing data and writing a paper. Any tips on how to work on the project efficiently? Any guides on how to approach research?

 

[adhominem]

what kind of research? Basic science? Epi? Chart review? Give us an idea of what you're actually doing and what kind of a team you're working with so we can actually give you useful non-generic advice.

[FailureToThrive]

26 minutes ago, adhominem said:

what kind of research? Basic science? Epi? Chart review? Give us an idea of what you're actually doing and what kind of a team you're working with so we can actually give you useful non-generic advice.

Clinical research and case reports, thanks!

[Albs]

.

[robclem21]

On 5/23/2018 at 12:22 AM, MD_2021 said:

I'll be working on a research project this summer. It will be my first time analyzing data and writing a paper. Any tips on how to work on the project efficiently? Any guides on how to approach research?

 

Strap yourself in, prepare for endless delays and frustration. Prepare to have your work scrutinized to no end. There is no such thing as efficiency. Its fun though. Enjoy, welcome to research.

[Bede]

I do basic science. I start with writing the paper followed by experiments. Your intro is your literature search followed by stating your hypothesis. This way you know what has/hasn't been done. Then write materials and methods. That's your experimental plan. Now start results. Obviously you don't have any but think of what data you want. Graphs, schematics, images, etc. Now do your experiments. As you get data, paste into your result section. Once you have all of the figures you planned on, put them into words and discuss. Voila- you have yourself a paper. 

Also see How to write a scientific paper by George Whitesides. 

[ExercMed]

On 6/22/2018 at 6:40 PM, Bede said:

I do basic science. I start with writing the paper followed by experiments. Your intro is your literature search followed by stating your hypothesis. This way you know what has/hasn't been done. Then write materials and methods. That's your experimental plan. Now start results. Obviously you don't have any but think of what data you want. Graphs, schematics, images, etc. Now do your experiments. As you get data, paste into your result section. Once you have all of the figures you planned on, put them into words and discuss. Voila- you have yourself a paper. 

Also see How to write a scientific paper by George Whitesides. 

This is how you bias your data to show what you want it to show... Clearly you want to do a literature review, have a hypothesis, and a rough experimental plan, but DO NOT "think of what data you want". You don't know what the data will show you until you start doing experiments. Those experiments probably wont tell you the full story, so then you plan to do more experiments based on that. Once you have data that you think tells a cohesive story, THEN you can start to write a paper around the data. Writing the paper first introduces MASSIVE bias as you don't know what you are going to see. 90% of the time, experiments don't work out or have insignificant results that don't merit further discussion (an unfortunate truth about our peer review system). If that's the case then your paper, that you spent all this time writing, is dead in the water and needs to be revised.

To quote the article by George Whitesides that the above poster suggested, "Realize that your objective in research is to formulate and test hypothesis, to draw conclusions from these tests, and to teach these conclusions to others. Your objective is not to 'collect data'." 

[Eudaimonia]

1 hour ago, ExercMed said:

This is how you bias your data to show what you want it to show... Clearly you want to do a literature review, have a hypothesis, and a rough experimental plan, but DO NOT "think of what data you want". You don't know what the data will show you until you start doing experiments. Those experiments probably wont tell you the full story, so then you plan to do more experiments based on that. Once you have data that you think tells a cohesive story, THEN you can start to write a paper around the data. Writing the paper first introduces MASSIVE bias as you don't know what you are going to see. 90% of the time, experiments don't work out or have insignificant results that don't merit further discussion (an unfortunate truth about our peer review system). If that's the case then your paper, that you spent all this time writing, is dead in the water and needs to be revised.

To quote the article by George Whitesides that the above poster suggested, "Realize that your objective in research is to formulate and test hypothesis, to draw conclusions from these tests, and to teach these conclusions to others. Your objective is not to 'collect data'." 

Absolutely- thank you for writing this. I was shocked reading that method. It may seem more appropriate/doable for a translational or clinical study maybe (I'm not familiar with that kind of research) but my experience with basic science has been far from knowing where the chips will land, unless it's purely an observational study and you are sure that you won't be interested in looking at additional parameters/effects. Cannot count the amount of times that new data completely changed the direction of a project, scrapping the original objective as something unexpected or more interesting came up. Like you said, you come up with new methods as you decide to investigate something further, so why not just leave the writing to the end. Only after seeing what all your data tells you can you write a nice introduction that includes relevant information. 

[Bede]

Sorry guys but you are completely wrong. (Also, I never said that I write the conclusions in advance, I said an intro/lit review, methods, and leave the results blank) What I am doing is starting with a hypothesis and collecting evidence to support it. I generally have an idea what I'm going to see before I start. That's what a hypothesis is. (I'm wrong lots too, but that's a different story). That's the way science is supposed to work. What you're suggesting is that you will look at the data and then formulate a hypothesis post hoc. This is completely wrong, although very common in many scientific fields. Unless I'm reading you wrong, your method is one of the reasons why the p-value is almost dead. When you start with, "let's see what the data shows", you are completely undermining the assumptions that go into any inference testing. 

There is a huge amount of literature available now on these fishing expeditions that pass as research these days... Start with John Ionnidis. 

In reply to the comment, "new data changed the direction of the study..." Very true. That's happened to me as well many times. What is supposed to happen in that case is that your results are now the preliminary results for a new study where you reformulate the hypothesis and re test it. It sounds to me though like you formulated a hypothesis, it was disproven by the results, you changed your hypothesis post hoc and did inference testing and published. That's NOT how science is supposed to work. 

[Eudaimonia]

With your elaborations I can see how it can apply to some types of research, for example an objective where you want to provide evidence of a phenotype existing. Then your experimental plan is not likely to change much. Indeed we start with a hypothesis, gather some preliminary data, and then refine the hypothesis to be more specific based on that new information. It seems you are caught up in the statistics of hypothesis testing, which I agree with in isolation, but I was mostly talking about scientific hypotheses. I don't let an insignificant result change my hypothesis 180 degrees if a trend is there. It will simply remain as insignificant if other completely different experiments are able to support that trend. No single experiment is going to warrant a conclusion. Of course, if I had insignificant results no matter the methods I used to test a hypothesis, I would not have a publishable paper, unfortunately (that's a different conversation). Your last impression about changing hypothesis is not what I do, since my overall conclusion doesn't hinge on the results of any single statistical test. I imagine it would be more so the case if you were trying to test "is this drug effective in ___" for EBM but I'm trying to delineate signaling pathways, hope you see the difference

[adhominem]

To be fair everyone, the OP did say they're writing case reports. Most of what you're discussing isn't even relevant to that kind of work.

[Eudaimonia]

18 minutes ago, adhominem said:

To be fair everyone, the OP did say they're writing case reports. Most of what you're discussing isn't even relevant to that kind of work.

Haha I know we derailed, but basic sciences was mentioned and caught me surprised

[Bede]

Eudaimonia,

I get what you're saying.

I'm probably a bit more anal about this sort of stuff now since I see so many fishing expeditions and I never really learned to do otherwise until well after my PhD. Most of us forget the very basics of the scientific method taught in grade 9 and get caught up in the "industry" of science, where every publication counts.

[TheSalmonMousse]

On 6/29/2018 at 7:25 AM, Bede said:

fishing expeditions that pass as research these days... Start with John Ionnidis.

I want to say something in defense of fishing expeditions, being that I'm somewhat of a fisherman myself. I think the main issue with them is the abundance of poorly conceived experiments and misuse of statistics . That criticism is welcome and necessary when it's deserved. Unfortunately, I've found that a lot of the criticism I've faced or been privy to all too often come from a more ideological place. I once heard a prof outright question the usefulness of screens and large-scale studies as he believed any results from these experiments could (and thus should!) be obtained through traditional reductionist hypothesis-driven methods.

I personally could not write a paper in way you prescribe. Not that it's not a perfectly valid approach, but I feel it leaves too little room for serendipity. What happens when your results bring you somewhere you hadn't anticipated? Also, there's the very real - and unfortunate - problem of what do you do with negative results.

[Dr Priya Verma]

Basically, Research depends on subjects of research, Best idea to research is taking reviews and get practical work in related fields.

[Mansi@30]

Its all depend on your subject. What you want to research and you can take a lot of information through the reviews.