Research methods, without the jargon.

Module 01What a hypothesis actually is.

A hypothesis is not a wish. It is not the summary of what your data ended up showing. It is a committed guess — a statement you write down before you look, saying what you expect to find and, crucially, why.

The commitment is the point. If you only write down your "hypothesis" after the results are in, you have not tested anything. You have described what happened and dressed it up in the language of science. That is a summary, not a hypothesis. Judges spot this instantly.

The working form is: If I do X, then Y will happen, because of Z. The "because" is the load-bearing word. Without it you have a prediction. With it, you have a mechanism — a claim about how the world works that your experiment can push back on.

Module 02Variables — the three kinds.

Every experiment has three kinds of variables. Naming them correctly is not a formality. It is how you find the mistake in your own design before a judge does.

Take a concrete case. You want to see whether your home WiFi is really slower further from the router. You measure download speed at 3 m, 6 m, and 9 m from the router, ten times at each distance, using speedtest.net on the same phone.

The independent variable is distance from the router (3, 6, 9 metres). That is the one thing you are deliberately changing. The dependent variable is download speed in Mbps — what you measure, what you expect to change in response. Your controlled variables are the ones you decide in advance to hold constant: same phone, same speedtest server, same time of day, same day of the week, same number of other devices connected to the WiFi, same doors open or closed. If your father starts a Netflix stream in the middle of your fifth reading, that is a controlled variable escaping your control — and you note it in the logbook.

Module 03Controls, and why they matter more than you think.

A control is the "no-treatment" version of your experiment. It answers the single most important question a judge will ever ask you: compared to what?

Without a control, you can describe what happened. You cannot claim it was because of your treatment. Suppose you make ten friends drink tulsi tea when they have sore throats. Eight report feeling better in two days. Is that the tulsi? Or would they have felt better anyway? You cannot say — you did not measure the alternative.

Now give ten more sore-throated friends a cup of plain warm water at the same temperature, in an identical cup, on the same schedule. If seven of them also feel better in two days, your tulsi story collapses. If only three do, you have evidence. The control does not just improve your experiment. It creates your ability to make a claim at all.

Module 04Sample size, repeats, and why one time is never enough.

One measurement is a story. Three is a finding. Ten starts to look like evidence. The reason is boring but non-negotiable: the world is noisy, and a single reading tells you almost nothing about the underlying truth.

Take your resting heart rate. Sit still, wait a minute, count for 60 seconds. Say you get 74. Is your true resting heart rate 74? Not necessarily. Count again five minutes later — you might get 71. Then 76. Then 72. Then 78. The average of those five (74.2) is a better estimate of your resting heart rate than any single reading was. That is what repeats buy you: a way to see past the noise.

For a school-level project, the rule of thumb is: minimum three repeats per condition. Five is safer. If you cross about thirty samples per group, you have earned the right to start using statistical tests seriously. Below three, do not draw a conclusion — you literally do not have enough information.

Module 05Basic statistics without the maths.

You do not need to become a statistician. You need to know four things well enough that you cannot be misled by them, and cannot mislead others.

Mean versus median. The mean is the average — add up the numbers and divide by how many. The median is the middle value when you line them all up. When the data is roughly symmetric they are close. When it is skewed, they diverge, and the median usually tells the truer story.

Consider household monthly income on your gali. Nine families earn between ₹20,000 and ₹60,000. One family — the businessman at the end — earns ₹15,00,000. The mean monthly income on your street is now above ₹1,80,000. That number is not wrong. It is just useless. The median, around ₹40,000, actually describes life on that street. Whenever one big value can distort the picture, prefer the median.

Spread. The standard deviation is a formal way of asking "how much do the numbers argue with each other?" A small standard deviation means your measurements clustered tightly. A large one means they scattered. If your five heart-rate readings were 74, 71, 76, 72, 78, they are close together — small spread. If they were 55, 92, 61, 88, 70, something is going on — large spread — and reporting the average alone would hide it.

Is this real, or could it be chance? When you compare two groups and their averages differ, some of that difference is real, and some is just noise. Statisticians summarise this in a single number called a p-value: roughly, the probability that a difference this big could have happened by pure luck if the two groups were actually the same. The convention is that if p is less than 0.05 — a one-in-twenty chance — you are allowed to call the difference "statistically significant". Under 0.05 does not mean "definitely true". It means "unlikely enough to be worth taking seriously".

The one test worth learning. For comparing the averages of two groups — treatment vs. control, near vs. far, tulsi vs. water — use a t-test. In Google Sheets, drop your two columns of data into two ranges and type =T.TEST(range1, range2, 2, 2). It returns a p-value. That is the test 90% of student projects need, and it lives inside a spreadsheet you already have.

Module 06Citing sources properly.

You cite for two reasons. First, honesty — when an idea is not yours, saying so is the minimum owed to the person whose it was. Second, verifiability — someone else should be able to open the exact source you used and check whether you read it correctly.

What needs citing? Any fact you did not measure yourself. Any figure, table, or graph you did not make yourself. Any argument or definition lifted — even paraphrased — from someone else's work. If in doubt, cite it. There is no penalty for over-citing. There is a real one for the opposite.

APA style is the safest default for STEM projects. The essentials: author, year, title, source. The three shapes you will use most:

Plagiarism, in one sentence, is presenting someone else's words, ideas, or figures as your own. It disqualifies at IRIS and ISEF. Period.

The six habits.

One line from each module. If you internalise nothing else, internalise these:

Where to go next.

Methods are only half the picture. These three pair well with what you just read.