【Math B】Statistical Inference — The Magic of Decoding Data Behind the Scenes
Hello! Welcome to the Math B "Statistical Inference" page.
When you hear the word "statistics," you might think, "The calculations look tedious..." or "The formulas seem hard..." But don't worry!
This chapter covers one of the most useful applications of math in the real world: "estimating the whole from a subset of data." This is the exact logic behind things like TV ratings and determining the effectiveness of new medicines.
While it is a frequently appearing topic on the Common Test, it can actually become a major source of points if you grasp the key points. Let's master it together!
1. Random Variables and Probability Distributions
First, let's organize the basic terminology. If these concepts are shaky, everything that follows will be difficult!
What is a Random Variable?
Like the numbers on a die, a variable whose value determines the probability is called a random variable (usually denoted as \(X\)).
Expectation, Variance, and Standard Deviation
These are numbers that represent the "characteristics" of data.
- Expectation \(E(X)\): This is the "mean." It represents what the average value would be if you repeated the experiment many times.
\(E(X) = x_1p_1 + x_2p_2 + \dots + x_np_n\) - Variance \(V(X)\): This represents the "dispersion" or spread of the data. A larger value means the data is more scattered.
\(V(X) = E(X^2) - \{E(X)\}^2\) (Remember it as: "Mean of squares" minus "square of the mean"!) - Standard Deviation \(\sigma(X)\): This is the square root of the variance. You can think of it as bringing the units back to their original state.
\(\sigma(X) = \sqrt{V(X)}\)
【Key Point】Transformation of Variables
When \(Y = aX + b\), the following relationships hold. This is an essential technique for simplifying calculations!
・\(E(aX + b) = aE(X) + b\) (The mean changes proportionally)
・\(V(aX + b) = a^2V(X)\) (Variance is multiplied by the square of \(a\). The constant \(b\) disappears because it doesn't affect the spread!)
・\(\sigma(aX + b) = |a|\sigma(X)\)
2. Binomial Distribution
This is the distribution used when you repeat a "success or failure" scenario multiple times.
When do we use it?
It is used when you "repeat an event with probability \(p\), \(n\) times," such as "how many times a ball goes into the hoop after 10 shots" or "how many times a '1' comes up after rolling a die 100 times." This is denoted as \(B(n, p)\).
How to remember the formulas
The expectation and variance for the binomial distribution \(B(n, p)\) are very simple formulas:
・Expectation: \(E(X) = np\)
・Variance: \(V(X) = np(1-p)\) (*Note: \(1-p\) is sometimes written as \(q\))
Example: If you take 100 free throws (\(n=100\)) with a success rate of 20% (\(p=0.2\))
How many do you think you'll make on average? \(100 \times 0.2 = 20\) shots, right? This is the expectation \(np\). It’s intuitive and easy to understand!
【Trivia】
When \(n\) is sufficiently large, the shape of the binomial distribution begins to look like the "normal distribution" we will explain next. Using this property to simplify calculations is a classic tactic on the Common Test.
3. Normal Distribution
The most important distribution in statistics: the "symmetrical bell-shaped curve."
Standardization: The Magic Spell \(Z = \frac{X - m}{\sigma}\)
There are many kinds of data with different means (\(m\)) and standard deviations (\(\sigma\)) in the world, making them hard to compare as they are. Therefore, we convert them so the "mean is 0 and the standard deviation is 1." This is called standardization.
【Common Mistake】
Be careful not to put \(\sigma - m\) in the numerator or use \(V(X)\) in the denominator for the standardization formula!
Remember the chant: "(Value - Mean) ÷ Standard Deviation."
How to use the Normal Distribution Table
Using the standardized \(Z\) value, you look up the probability in the "Normal Distribution Table" at the back of your textbook.
・\(P(0 \leqq Z \leqq 1) \approx 0.3413\)
・\(P(0 \leqq Z \leqq 1.96) \approx 0.475\)
* The number \(1.96\) will play a huge role in "Estimation" later, so make friends with it!
4. Population and Sample
Now we get to the real "inference" part!
・Population: The entire group being studied (e.g., all high school students in Japan)
・Sample: The portion actually examined (e.g., 100 randomly selected high school students)
Properties of the Sample Mean \(\bar{X}\)
This is the trickiest part, but if you get through this, you're almost at the finish line!
Let the population mean be \(m\) and the standard deviation be \(\sigma\). For a "sample mean \(\bar{X}\)" taken from \(n\) samples:
- Expectation \(E(\bar{X}) = m\) (The mean of the sample means is the same as the original mean!)
- Variance \(V(\bar{X}) = \frac{\sigma^2}{n}\) (The larger the sample size, the smaller the dispersion of the mean!)
- Standard Deviation \(\sigma(\bar{X}) = \frac{\sigma}{\sqrt{n}}\) (Sigma divided by root \(n\)!)
【Analogy】
Imagine tasting soup. One ladle (sample) should taste roughly the same as the entire pot (population). The more you taste (the larger \(n\) is), the more accurate your assessment of the overall taste becomes (less dispersion).
5. Estimation
We predict the true value of the population (population mean \(m\)) from sample data.
Estimation of Population Mean (95% Confidence Interval)
This is the formula to say, "The true mean \(m\) is roughly within this range!"
Using the sample mean \(\bar{X}\), the probability that \(m\) falls in the following range is 95%:
\(\bar{X} - 1.96 \frac{\sigma}{\sqrt{n}} \leqq m \leqq \bar{X} + 1.96 \frac{\sigma}{\sqrt{n}}\)
【Key Point】
・For a 95% confidence interval, use 1.96
・For a 99% confidence interval, use 2.58
These numbers are always specified in the problem statement, so stay calm and plug them in.
6. Hypothesis Testing
This is a mathematical way to judge suspicions like, "Isn't this die rigged?"
Learn the Steps
- Set a Hypothesis (Null Hypothesis): Assume that "this die is fair (no bias)."
- Calculate the Probability: Calculate the probability of the current result happening if the hypothesis were true.
- Judge: If that probability is smaller than a predetermined threshold (significance level, usually 5% or 1%), conclude, "It's too strange for this to happen by chance! The hypothesis must be wrong!" (This is called rejection).
【Analogy】
It's just like a trial. You start with the hypothesis that "the defendant is innocent," and if you find evidence that is too unnatural to ignore, you conclude that "it's impossible to claim they are innocent (i.e., they are guilty)."
Summary: Tips for Cracking the Common Test
It might feel difficult at first, but don't worry. Just perfect these three points:
- Memorize \(np\) and \(npq\) for the binomial distribution.
- Be able to use the standardization formula \(Z = \frac{X - m}{\sigma}\) automatically.
- Remember the standard deviation of the sample mean is \(\frac{\sigma}{\sqrt{n}}\) (don't forget to divide by \(\sqrt{n}\)!).
Statistical inference relies on fixed calculation patterns. If you write the formulas on a piece of paper, tape them to your wall, and solve a few basic problems, you'll get the hang of it in no time. I'm rooting for you!