What is Likelihood?

S3-SA3-0173

Grade Level:

Class 9

AI/ML, Data Science, Physics, Economics, Cryptography, Computer Science, Engineering

Definition

What is it?

Likelihood measures how well a specific statistical model or hypothesis explains observed data. It's not a probability itself, but a way to compare different explanations for the same event.

Simple Example

Quick Example

Imagine you have a bag of marbles, and you don't know if it has mostly red or mostly blue marbles. You pull out 5 marbles and get 4 red and 1 blue. The 'likelihood' tells you which scenario (mostly red or mostly blue) is more probable given what you just saw.

Worked Example

Step-by-Step

Let's say a cricket team usually wins 70% of its matches. However, in the last 10 matches, they won only 3. We want to find the likelihood of this happening if their true win rate is still 70% versus if their true win rate has dropped to 30%.

---Step 1: Define the scenarios (hypotheses).
Scenario A: True win rate is 70% (p = 0.7).
Scenario B: True win rate is 30% (p = 0.3).

---Step 2: Define the observed data.
They won 3 out of 10 matches (k = 3, n = 10).

---Step 3: Use the binomial probability formula P(k successes in n trials) = C(n, k) * p^k * (1-p)^(n-k).

---Step 4: Calculate Likelihood for Scenario A (p=0.7).
L(p=0.7 | 3 wins in 10) = C(10, 3) * (0.7)^3 * (0.3)^7
C(10, 3) = 10! / (3! * 7!) = (10*9*8) / (3*2*1) = 120
L(p=0.7 | 3 wins in 10) = 120 * (0.343) * (0.0002187) = 120 * 0.00007507 = 0.0090084

---Step 5: Calculate Likelihood for Scenario B (p=0.3).
L(p=0.3 | 3 wins in 10) = C(10, 3) * (0.3)^3 * (0.7)^7
L(p=0.3 | 3 wins in 10) = 120 * (0.027) * (0.0823543) = 120 * 0.0022235661 = 0.266827932

---Step 6: Compare the likelihoods.
Likelihood for Scenario A (0.009) is much smaller than Likelihood for Scenario B (0.267).

---Answer: The observed data (3 wins in 10) is much more likely if the team's true win rate has dropped to 30% than if it remained at 70%.

Why It Matters

Likelihood is a core idea in AI/ML, helping computers 'learn' from data by finding the best models. Data scientists use it to build predictive models, engineers use it to design efficient systems, and economists use it to understand market trends. It's crucial for careers in AI development, data analytics, and scientific research.

Common Mistakes

MISTAKE: Thinking likelihood is the same as probability. | CORRECTION: Probability refers to the chance of observing data given a fixed model. Likelihood refers to how well a model fits observed data. They are related but distinct concepts.

MISTAKE: Assuming a higher likelihood means a model is 'correct'. | CORRECTION: A higher likelihood simply means the model explains the observed data better than other models being compared. It doesn't guarantee the model is the absolute truth.

MISTAKE: Using likelihood values directly as probabilities for models. | CORRECTION: Likelihood values themselves are not probabilities of the model being true. They are relative measures for comparison. To get model probabilities, you'd need to use Bayesian methods.

Practice Questions

Try It Yourself

QUESTION: You flip a coin 5 times and get 4 heads. What is the likelihood of this outcome if the coin is fair (probability of heads = 0.5)? | ANSWER: Using P(k successes) = C(n, k) * p^k * (1-p)^(n-k), where n=5, k=4, p=0.5. C(5,4) = 5. So, 5 * (0.5)^4 * (0.5)^1 = 5 * 0.0625 * 0.5 = 0.15625

QUESTION: A factory produces LED bulbs. Historically, 2% are defective. In a batch of 100 bulbs, 5 are found defective. Calculate the likelihood of observing 5 defective bulbs if the defect rate is still 2%. (Hint: Use binomial probability) | ANSWER: n=100, k=5, p=0.02. L = C(100, 5) * (0.02)^5 * (0.98)^95. C(100, 5) = 75,287,520. (0.02)^5 = 3.2e-9. (0.98)^95 approx 0.146. L approx 75,287,520 * 3.2e-9 * 0.146 approx 0.035

QUESTION: You observe 2 rainy days in a week (7 days). What is the likelihood of this happening if the probability of rain on any given day is 0.3? What if the probability of rain is 0.5? Which scenario has higher likelihood? | ANSWER: For p=0.3: L = C(7,2) * (0.3)^2 * (0.7)^5 = 21 * 0.09 * 0.16807 = 0.3176. For p=0.5: L = C(7,2) * (0.5)^2 * (0.5)^5 = 21 * 0.25 * 0.03125 = 0.164. The scenario with rain probability 0.3 has a higher likelihood (0.3176 vs 0.164).

MCQ

Quick Quiz

Which of the following best describes 'likelihood'?

The chance of an event happening.

How well a model explains observed data.

The certainty of a prediction.

The total number of possible outcomes.

The Correct Answer Is:

Likelihood quantifies how plausible a particular statistical model or hypothesis is, given the data that has been observed. It's about evaluating the model, not predicting the event or counting outcomes.

Real World Connection

In the Real World

Imagine your favourite food delivery app like Zomato or Swiggy. When they recommend restaurants or dishes to you, they use algorithms that calculate the 'likelihood' that you will like a certain item based on your past orders and ratings. Similarly, in weather forecasting, meteorologists use likelihood models to determine which climate model best fits the current atmospheric data to predict rain or sunshine.

Key Vocabulary

Key Terms

HYPOTHESIS: An idea or explanation for something based on known facts but not yet proven. | MODEL: A simplified representation of a system or process. | DATA: Facts and statistics collected together for reference or analysis. | BINOMIAL PROBABILITY: The probability of getting a certain number of successes in a fixed number of trials.

What's Next

What to Learn Next

Next, you can explore 'Maximum Likelihood Estimation (MLE)'. This concept builds on likelihood to find the best possible parameters for a model that maximize its likelihood given the observed data. It's a key technique used in many advanced fields!