Basics of a Good Research Article

Ok, so I ideally want to be doing one of these blogs a week, but somehow ended up really busy for a few weeks, I apologize.


This week we are exploring what makes a study good or bad. Scientific research papers can be pretty complicated to read and understand, but there are some basics you can use to determine if what you are reading has any weight. Keep in mind I am only using the very basics of what makes a good study, there are some things that can be tweaked and altered and your study will still be fine but for the sake of this blog we'll be using some quick and easy stuff to pick out.

First and foremost any research study you look at should be a peer-reviewed journal article. This means the article in question was looked at by a group of experts and peers in the field and determined to be an all around good study. This doesn't mean that the study is flawless, but it does help it hold weight in the real world.

The Hypothesis

I'll start out with one we are familiar with. The first thing you want to look at is whether the researcher chose a hypothesis that is testable. It is simply not testable at this point to state that you hypothesize that there is a planet inhabited by green people orbiting a distant star. As I have pointed out before there is simply no way to prove this.Typically when you are reading a research article a researcher will state what is known as the null hypothesis. The null hypothesis is the assumption that there is no relationship between your the variables you are testing and that any observed patterns are due merely to chance. This is important when it comes to statistical testing as a researcher will be attempting to show that there really is a difference between the data sets and they will be trying to reject or disprove the null hypothesis.  The purpose of the null is to allow you to compare your hypothesis against something else. An example might be:

Hypothesis: Watering plants will help them to grow bigger.

Null Hypothesis: There will be no difference in growth of the plants due to water.

The researcher's study will attempt to determine whether or not they are able to reject the null (the plants indeed grew) or accept the null (whether the plants were watered or not made no difference).

Sample size

In general, the larger the sample a researcher is able to get the more data they will have on hand to work with and analyze. The number I use as a rule of thumb is a sample size of 30 or more. Below that, it becomes hard to determine whether or not your results can be generalized to the rest of the population. Studies will use smaller sample sizes quite frequently and effectively, but the study is usually an initial one which requires more research. Sometimes a larger sample is just not possible to find, especially if you are researching a relatively rare phenomenon. For general purposes though I would go with the 30 or more rule.

Placebos

Whenever doing a study with human and animal participants it is important to have a placebo group; this is the group that is given no treatment and will help researchers be able to tell whether or not their results are due to chance. An example of this would be if you are testing the effectiveness of a headache pill. The pill given to one group will be the actual pill for headaches, the pill given to another group will be a sugar pill which should have no effect. Aside from this, both pills should look exactly the same. This is particularly important when dealing with medical research, as oftentimes patients will react to a treatment simply because it is a medical intervention and they expect that it will improve their condition.

Double Blind

We are very influenced by our surroundings, a lot of the time without even knowing it. In order to control for this in an experiment, researchers will use a double blind study. This means that the participants of the study don't know which group they are in (placebo or test group) and the researchers are also blind to what group the participants are in. This helps to ensure that the people carrying out the testing are not influencing their participants, and the participants are not influencing the study by knowing which group they are in.

P-value

This is a more confusing part of research studies- when you get to the statistics, how can you tell whether or not the study you are researching had a significant result? One way to determine this is in the P-value given for the study. The P-value is the probability that your results are due entirely to chance. Most researchers will use a value of 0.05 representing a 5% chance that the results were due to chance alone and not the effect/treatment you are studying.  If a p-value comes up over 0.05 then you would fail to reject your null hypothesis (this means your study showed no effect) if it is below 0.05 then you can reject the null hypothesis (This means the effect in your study was not caused just by chance). This value can be adjusted, but as a general rule of thumb a researcher will use 0.05 or lower and what you should be on the look out for when reading a research article is that they've increased their p-value to greater than 0.05.

Meta-analysis

This isn't generally a part of one research study, but collects the data from many different research studies with similar hypotheses and combines them to see if there is a common trend. An example would be a meta-analysis study of homeopathy; a couple of studies may have shown a weak positive result, but the combination of many sets of data show a strong negative.


Now you should have a better idea of what to look for when you are checking out various information on the internet.  I would however remember that I have only laid out the very basics of a research article, there is much more that goes into making a study good or not, so if you find yourself on the fence about an article I suggest finding a sciencey friend to look it over and see what they think!