24 March 2018

How to read a scientific paper 2

This artist’s impression shows how the Milky Way galaxy would look seen from almost edge on and from a very different perspective than we get from the Earth. The central bulge shows up as a peanut shaped glowing ball of stars and the spiral arms and their associated dust clouds form a narrow band.

Another apology is needed for this post. I was absent from my regular social media sharing of science news. Sorry about that. I was feeling a bit under the weather.
So let us continue with tips on how to read scientific papers.
In the previous post, I chose an article Discovery of Gamma-Ray Emission from the X shaped Buldge of the Milky Way. You can find a full article for free at arxiv.org. And if you feeling posh, you can also buy the version that was published in Nature.
BTW. scientists love to publish papers in Nature and quote them, but not to read them for serious study. The Nature papers are extremely short, leaving out loads of important details scientists need to reproduce the analytics methods, so we wait a bit for another version of the same paper to appear in different journal, more specialized one.
Anyway, back to the reading. In a previous post, I mentioned a ‘triage’ average grad student uses when reading scientific papers, and I will follow that ‘triage’ here.
1. Read an abstract.
2. Look at the images and read captions
3. Read conclusion/summary
4 Read the paper in details.
So, the second step is looking at the images.


Paper has colored images, 11 in total. So let us go one by one. Again, due to copyrights, I cannot use any images or parts of papers in my article, so you’ll have to download it for yourself and just follow.

Figure 1

First one is on page 3 of the paper. Figures usually have detailed captures explaining them, because a majority of grad students follow this triage.
So when you look at the images, first read the capture. It explains a figure a lot. This particular figure has two panels, bigger one main, and the smaller one zoomed up part of the bigger panel, the one that looks like a mess in the bigger panel.
I can see why the authors zoomed the part of the frame, without zoom in one cannot really see what is going on.
So capture mentions residuals, however, it does not say residuals of what. The answer is in the paper, so that is not a problem at the moment. Use of residuals means, that the raw data, as they are, could not show any change, so the original parameter value is manipulated to enhance tiny variations so that we can see what is going on there. This technique is often used all across the science to detect the changes that are subtle.
For example, all of us know the value of the Earth gravity, and we know it does not noticeably change when we go around the planet. But the fact is it does changes, in very small amounts. To see those amounts clearly, residuals are used. The average value of the Earth gravity is removed from the data and suddenly barely visible variations then became glaringly obvious.
Something similar was done here. The zoomed part shows something that seems like confirmation of the result. The part with the dotted circle shows the part that does not confirm the result, but capture indicates there might be an explanation.

Figure 2

This one is a graph. A fancy one. Scientists love graphs, and often in physics, we will use some kind of equation to get graphs that seem simple. You can see in y-axis the value used is not simple but looks like a part of the equation. This technique, using parts of the equation as y-axis value makes the resulting plot almost linear.
In physics, this is done on a regular basis, because all simple physics was discovered a long time ago, leaving only super-complicated stuff for today. Humans tend to comprehend relationships only if they are presented in simple curves, so physicists do this little manipulation, making sure the parts of the equations used are well proven one so that only new relationship is shown in its’s virtual simplicity.
To me, this figure is neat evidence for the claim of the paper. But, still, I have no clue about the details, so onward to another figure.

Figure 3

This figure is in an appendix. Remember what I told you that the Nature papers are short? Well, this appendix contains all the details that had to be removed from the Nature paper but are usually necessary for useful scientific papers.
Figure 3 shows a comparison of the two methods. Remember, the abstract was stating that a new method was used. And here, in this figure, we suppose to see why. Two left columns show way more details than two right columns. The capture does not say which columns belong to which model, but my guess is two left columns are the new model. And yep, I see why is better, the resolution of whatever they are watching is way better on those, allowing us to see the details.

Figure 4

This one is a complete mystery to me. The caption is not really telling anything, all I can guess from here is that both hydrodynamic and interpolated methods give quite similar results for dust. But why is this important I have no idea? Reading the paper in details should give the answers.

Figure 5

Another mysterious figure. All I can get from this one is that it is most likely used to illustrate methods used in the paper. But why is important, the paper will tell.

Figure 6

Ok, this one is clear, it supposed to show this X-bulge. Now I have to admit, that I have no clue what is X-bulge. So I googled it to see what is it. I found this article and scrolled down to the image that shows another form of residuals, what is left over of our galaxy after we subtract the model of a symmetrical bulge. And yep, it is some kind of X shape. https://www.inverse.com/article/18507-milky-way-x-bulge
So now, this figure sort of makes sense. It supposed to mark where those extra stars are located in our galaxy.

Figure 7

And now we see a template for the nuclear bulge. Ok. They either did this to test their techniques or it was needed for analysis. Paper will explain. (Btw. the nuclear bulge is that bump in the center of the galaxy where it is way thicker.)

Figure 8

Now we’re talking. This figure is showing the newly identified sources and how much of a possible error is connected with them. Brighter color means less error. And now we see why they need to explain that supernova remnant, the significance there is higher, and yet, that one is not a part of x bulge. Definitely, something that might f-k up their conclusion, if not explained.
This is actually a signal of the good paper because authors do not hide a rather inconvenient result. Every study has inconvenient results because that’s how Nature works, so when you see those in a paper, the probability that paper is good jumps up.

Figure 9

Another almost-linear plot. At this point, I will note that the majority of the points seem to be in one line, a good sign. And a few points outside line means the authors did not lie about the results.

Figure 10

Illustration of another must during the introduction of the new method. An example of using the new method to get the old results. If the new method cannot produce old results, then it is not a good method. I am not familiar enough with previous results to know immediately did they got it or not, but I’m comforted that there is an inclusion of the new method test.

Figure 11

Graph representing those residuals from the first two figures in way more details. Another figure to understand later with a reading of the paper.

Summary for images

So Figures showed that this paper is good (not surprising, it was published in Nature), but I did point why the figures showed it is a good paper. There is loads of stuff I do not get at the moment, but I’m ready to go to the next step. Reading the end of the paper.

No comments:

Post a Comment