Linear Transformation: A transformation of the form y = ax + b where a and b are constants
Additive Transformation: When b ≠ 0 and a = 1 (y = x + b)
Location Change: Adding/subtracting constants shifts the entire distribution
- Center measures: Affected by addition/subtraction (mean, median)
- Spread measures: Not affected by addition/subtraction (std dev, IQR)
- Shape: Not affected by addition/subtraction
New mean = Original mean + 10 = 25 + 10 = 35
New median = Original median + 10 = 24 + 10 = 34
New standard deviation = Original standard deviation = 5 (unchanged)
New IQR = Original IQR = 8 (unchanged)
Shape remains the same, skewness unchanged
New median = 34
New std dev = 5
New IQR = 8
After adding 10 to each value: new mean = 35, new median = 34, standard deviation = 5 (unchanged), IQR = 8 (unchanged). Adding a constant shifts the center but does not affect the spread or shape of the distribution.
• Addition effect: Mean, median increase by added constant
• No spread change: Standard deviation, IQR unchanged
• Shape preservation: Skewness and shape remain the same
Multiplicative Transformation: When a ≠ 1 and b = 0 (y = ax)
Scale Change: Multiplying by a constant changes the scale of the distribution
- Center measures: Multiplied by the constant (mean, median)
- Spread measures: Multiplied by the absolute value of the constant (std dev, IQR)
- Shape: Generally preserved (skewness unchanged)
New mean = Original mean × 2 = 12 × 2 = 24
New median = Original median × 2 = 11 × 2 = 22
New standard deviation = Original std dev × 2 = 3 × 2 = 6
New IQR = Original IQR × 2 = 5 × 2 = 10
Shape remains the same, skewness unchanged
New median = 22
New std dev = 6
New IQR = 10
After multiplying each value by 2: new mean = 24, new median = 22, standard deviation = 6, IQR = 10. Multiplying by a constant scales both center and spread measures by that constant, but preserves the shape and skewness of the distribution.
• Multiplication effect: Mean, median multiplied by constant
• Spread scaling: Standard deviation, IQR multiplied by absolute value of constant
• Shape preservation: Multiplication preserves relative relationships
Combined Linear Transformation: y = ax + b where both a and b are non-zero
Sequential Effects: First scale by a, then shift by b
- Center measures: Multiply by a, then add b (mean, median)
- Spread measures: Multiply by |a| (std dev, IQR)
- Shape: Preserved under linear transformations
New mean = (Original mean × 2) + 5 = (40 × 2) + 5 = 80 + 5 = 85
New median = (Original median × 2) + 5 = (38 × 2) + 5 = 76 + 5 = 81
New standard deviation = Original std dev × |2| = 8 × 2 = 16
New IQR = Original IQR × |2| = 12 × 2 = 24
Adding 5 doesn't affect spread; multiplying by 2 affects both center and spread
New median = 81
New std dev = 16
New IQR = 24
After applying y = 2x + 5: new mean = 85, new median = 81, standard deviation = 16, IQR = 24. The multiplication by 2 scales both center and spread, while the addition of 5 shifts only the center measures.
• Linear transformation: y = ax + b affects center by ax + b
• Spread effect: Only affected by multiplication (×|a|)
• Order matters: Scale first, then shift
Logarithmic Transformation: y = log(x), compresses large values more than small values
Skewness Reduction: Often makes right-skewed distributions more symmetric
- Shape: Reduces right skewness, makes distribution more symmetric
- Center: Both mean and median decrease, but relationship may change
- Spread: Typically reduces variability
Right-skewed distribution becomes more symmetric
Large values are compressed more than small values
Mean and median both decrease
Log transformation reduces the impact of large values
Standard deviation typically decreases
Values become more clustered around the center
Center: Both decrease
Spread: Likely decreases
Skewness: Reduced
After log transformation: The distribution becomes more symmetric with reduced right skewness. Both mean and median decrease, and the standard deviation typically decreases as well. The log transformation compresses the larger values more than the smaller values, making the distribution more compact and reducing the influence of outliers.
• Log effect: Compresses large values more than small values
• Skewness reduction: Particularly effective for right-skewed data
• Non-linear: Does not preserve linear relationships
Power Transformation: y = x^p where p is a power (e.g., square root when p = 0.5)
Root Transformation: A type of power transformation that reduces skewness
- Shape: Root transformations reduce right skewness
- Center: Both mean and median decrease but maintain relative ordering
- Spread: Values become more compressed toward center
New mean ≈ √2000 ≈ 44.7 (but not exactly, as mean of transformed ≠ transformed mean)
New median = √1500 ≈ 38.7 (median transforms exactly)
Right-skewed distribution becomes less skewed
Large values are reduced more than small values
Distribution becomes more compact
Distance between values decreases
Shape: Less skewed
Spread: More compact
Values: Compressed
After square root transformation: The distribution becomes less right-skewed as large values are compressed more than small values. The median transforms exactly (√1500 ≈ 38.7), and the distribution becomes more symmetric with values clustering closer to the center. The transformation makes the data more suitable for analysis requiring normality.
• Root effect: Reduces skewness, compresses large values
• Median preservation: Median transforms exactly
• Mean change: Mean changes but not in a simple way
Center: New = a(old) + b
Spread: New = |a|(old)
Shape: Preserved
Linear Transformations: Of the form y = ax + b, where a and b are constants. These transformations preserve the shape of the distribution while changing location and scale.
Non-linear Transformations: Such as logarithmic, square root, or power transformations. These can change the shape of the distribution, often used to reduce skewness.
Effect Patterns: Linear transformations affect center and spread systematically, while non-linear transformations primarily affect shape and can make distributions more symmetric.
- Identify transformation type: Linear (y = ax + b) or non-linear (log, root, power)
- Determine effect on center: Mean, median, mode
- Determine effect on spread: Standard deviation, IQR, range
- Determine effect on shape: Skewness, symmetry, modality
• Linear transformation: y = ax + b affects center by ax + b, spread by |a|, shape preserved
• Addition/subtraction: Only affects center measures
• Multiplication/division: Affects both center and spread
• Non-linear: Can change shape, skewness, and relationships