Shifting a Random Variable: Adding a constant to a random variable shifts its entire distribution by that constant. For Y = X + c: E[Y] = E[X] + c and Var(Y) = Var(X).
- Expected value shifts by the constant: E[X + c] = E[X] + c
- Variance remains unchanged: Var(X + c) = Var(X)
- Standard deviation remains unchanged: SD(X + c) = SD(X)
For Y = X + c: E[Y] = E[X] + c. Therefore, E[X + 3] = E[X] + 3 = 5 + 3 = 8.
Adding a constant doesn't change the spread of the distribution: Var(X + c) = Var(X). Therefore, Var(X + 3) = Var(X) = 4.
SD(X + 3) = √Var(X + 3) = √4 = 2.
E[X + 3] = 8 and Var(X + 3) = 4.
• Expected value shifting: E[X + c] = E[X] + c
• Variance invariance: Var(X + c) = Var(X)
• Location shift: Adding constant shifts location but not spread
Scaling a Random Variable: Multiplying a random variable by a constant scales both its location and spread. For Y = aX: E[Y] = aE[X] and Var(Y) = a²Var(X).
For Y = aX: E[Y] = aE[X]. Therefore, E[3X] = 3E[X] = 3 × 2 = 6.
Multiplying by a constant scales variance by the square of the constant: Var(aX) = a²Var(X). Therefore, Var(3X) = 3²Var(X) = 9 × 9 = 81.
SD(3X) = √Var(3X) = √81 = 9.
E[3X] = 6 and Var(3X) = 81.
• Expected value scaling: E[aX] = aE[X]
• Variance scaling: Var(aX) = a²Var(X)
• Spread scaling: Variance scales by the square of the multiplier
Linear Transformation: For Y = aX + b, the expected value is E[Y] = aE[X] + b and the variance is Var(Y) = a²Var(X). The constant b affects the mean but not the variance.
For Y = aX + b: E[Y] = aE[X] + b. Therefore, E[2X - 7] = 2E[X] - 7 = 2 × 4 - 7 = 8 - 7 = 1.
For Y = aX + b: Var(Y) = a²Var(X). The constant b doesn't affect variance. Therefore, Var(2X - 7) = 2²Var(X) = 4 × 5 = 20.
SD(2X - 7) = √Var(2X - 7) = √20 = 2√5 ≈ 4.472.
E[2X - 7] = 1 and Var(2X - 7) = 20.
• Linear transformation: E[aX + b] = aE[X] + b
• Variance scaling: Var(aX + b) = a²Var(X)
• Constant shift: Adding constant affects mean but not variance
Shifting: Adding a constant to a random variable, Y = X + c. This moves the entire distribution without changing its shape or spread.
Scaling: Multiplying a random variable by a constant, Y = aX. This stretches or compresses the distribution.
Linear Transformation: A combination of shifting and scaling, Y = aX + b. This changes both location and scale.
Expected Value Properties: E[aX + b] = aE[X] + b. Linear transformations preserve linearity.
Variance Properties: Var(aX + b) = a²Var(X). The constant term doesn't affect variance.
- Identify the transformation: Determine if it's shifting, scaling, or linear
- Apply expected value property: Use E[aX + b] = aE[X] + b
- Apply variance property: Use Var(aX + b) = a²Var(X)
- Calculate standard deviation: SD = √Var if needed
- Verify results: Check if the results make sense
- Interpret: Understand the practical meaning
• Linear transformation: E[aX + b] = aE[X] + b
• Variance scaling: Var(aX + b) = a²Var(X)
• Standard deviation scaling: SD(aX + b) = |a|SD(X)
• Shifting: E[X + c] = E[X] + c, Var(X + c) = Var(X)
• Scaling: E[aX] = aE[X], Var(aX) = a²Var(X)
Linear Transformation Application: Converting temperature from Celsius to Fahrenheit follows the linear transformation F = (9/5)C + 32. Apply transformation properties to find new mean and standard deviation.
The conversion from Celsius to Fahrenheit is: F = (9/5)C + 32, where a = 9/5 and b = 32.
E[F] = (9/5)E[C] + 32 = (9/5)(20) + 32 = 36 + 32 = 68°F.
SD(F) = |a|SD(C) = |9/5|SD(C) = (9/5)(5) = 9°F.
Var(F) = (9/5)²Var(C) = (81/25)(25) = 81°F².
The mean temperature in Fahrenheit is 68°F and the standard deviation is 9°F.
• Linear transformation: E[aX + b] = aE[X] + b
• Standard deviation scaling: SD(aX + b) = |a|SD(X)
• Real-world application: Temperature conversion
Composition of Transformations: When multiple transformations are applied sequentially, we can either work step-by-step or find the overall transformation and apply it directly.
E[Y] = 3E[X] - 5 = 3(10) - 5 = 25. Var(Y) = 3²Var(X) = 9(16) = 144.
E[Z] = -2E[Y] + 7 = -2(25) + 7 = -50 + 7 = -43. Var(Z) = (-2)²Var(Y) = 4(144) = 576.
Z = -2Y + 7 = -2(3X - 5) + 7 = -6X + 10 + 7 = -6X + 17. E[Z] = -6(10) + 17 = -60 + 17 = -43. Var(Z) = (-6)²Var(X) = 36(16) = 576. ✓
E[Z] = -43 and Var(Z) = 576.
• Step-by-step approach: Apply transformations sequentially
• Composition property: Transformations can be combined algebraically
• Verification: Both approaches yield the same result
Shifting: Adding a constant to a random variable, Y = X + c. This moves the entire distribution without changing its shape or spread.
Scaling: Multiplying a random variable by a constant, Y = aX. This stretches or compresses the distribution.
Linear Transformation: A combination of shifting and scaling, Y = aX + b. This changes both location and scale.
Expected Value Properties: E[aX + b] = aE[X] + b. Linear transformations preserve linearity.
Variance Properties: Var(aX + b) = a²Var(X). The constant term doesn't affect variance.
Standard Deviation Properties: SD(aX + b) = |a|SD(X). The absolute value ensures non-negativity.
- Identify the transformation: Determine if it's shifting (X + c), scaling (aX), or linear (aX + b)
- Apply expected value property: Use E[aX + b] = aE[X] + b
- Apply variance property: Use Var(aX + b) = a²Var(X)
- Calculate standard deviation: SD = √Var if needed
- Handle compositions: Apply transformations sequentially or combine algebraically
- Verify results: Check if the results make practical sense
• Linear transformation: E[aX + b] = aE[X] + b
• Variance scaling: Var(aX + b) = a²Var(X)
• Standard deviation scaling: SD(aX + b) = |a|SD(X)
• Shifting: E[X + c] = E[X] + c, Var(X + c) = Var(X)
• Scaling: E[aX] = aE[X], Var(aX) = a²Var(X)
• Composition: If Y = aX + b and Z = cY + d, then Z = caX + cb + d
• Simple: Pure shifting (X + c)
• Intermediate: Pure scaling (aX)
• Advanced: Linear transformation (aX + b)
• Complex: Composition of transformations
• Shifting affects mean only: E[X + c] = E[X] + c, Var(X + c) = Var(X)
• Scaling affects both: E[aX] = aE[X], Var(aX) = a²Var(X)
• Linear: E[aX + b] = aE[X] + b, Var(aX + b) = a²Var(X)
• Variance always scales by the square of the multiplier
• Standard deviation scales by the absolute value of the multiplier
Questions & Answers
Question: Why does the constant in variance transformation get squared?
Answer: Variance measures squared deviations from the mean:
- Var(X) = E[(X - E[X])²]
- When we transform Y = aX, the deviations become (aX - aE[X]) = a(X - E[X])
- So Var(Y) = E[(a(X - E[X]))²] = E[a²(X - E[X])²] = a²E[(X - E[X])²] = a²Var(X)
The squaring in the variance definition causes the multiplier to be squared.
Question: Why do we use absolute value for standard deviation scaling?
Answer: Standard deviation is always non-negative:
- SD(X) = √Var(X), so it's always ≥ 0
- If a is negative, then SD(aX) = √Var(aX) = √(a²Var(X)) = √(a²)√Var(X) = |a|SD(X)
- For example, if a = -2, then SD(-2X) = |-2|SD(X) = 2SD(X), not -2SD(X)
The absolute value ensures standard deviation remains positive.
Question: Do these transformation properties work for non-linear transformations like X² or log(X)?
Answer: No, these simple linear transformation properties only apply to linear transformations of the form aX + b:
- For non-linear transformations: E[g(X)] ≠ g(E[X]) in general
- Example: E[X²] ≠ (E[X])² unless Var(X) = 0
- For X²: Need to use E[X²] = Var(X) + (E[X])²
- For log(X): No simple formula exists; requires more advanced techniques
Linear transformations have special properties that make them particularly tractable.