aX+b

What $aX+b$ means

• Linear transformation: $Y=aX+b$ means every value of $X$ is multiplied by the constant $a$, then shifted by the constant $b$. This creates a new random variable $Y$ whose probabilities come from the same underlying outcomes.
• Role of constants: The constant $a$ controls scaling and possible reflection, while $b$ controls horizontal or vertical location shift in value space. This separation helps predict whether the center, spread, or both will change.

Core formulas to know

• Expectation rule: > Key formula: $E(aX+b)=aE(X)+b$. This works because expectation is linear, so scaling and adding constants distribute directly through $E(\cdot)$.
• Variance rule: > Key formula: $Var(aX+b)=a^2Var(X)$. The shift $b$ disappears from variance because variance measures spread around the mean, not absolute

Why expectation transforms linearly

• Weighted-average logic: Expectation is a probability-weighted average, so multiplying all outcomes by $a$ multiplies the average by $a$. Adding $b$ to all outcomes increases the average by exactly $b$.
• Operator viewpoint: Linearity gives $E(c_1X+c_2)=c_1E(X)+c_2$ for constants $c_1,c_2$. This principle is foundational across probability, statistics, and linear models.

Why variance uses $a^2$

• Spread around center: Variance is based on squared deviations, $Var(X)=E[(X-E(X))^2]$. When outcomes are scaled by $a$, deviations scale by $a$, and squaring produces $a^2$.
• Translation invariance: Adding $b$ shifts both $X$ and its mean by the same amount, so deviations from the mean are unchanged. Therefore $Var(X+b)=Var(X)$ holds for any constant $b$.

Step-by-step workflow

• Step 1: Standardize the expression: Rewrite forms like $X-c$ as $X+(-c)$ and $\frac{X}{k}$ as $\left(\frac{1}{k}\right)X$. This prevents sign mistakes and makes $a$ and $b$ explicit.
• Step 2: Apply formulas directly: Compute mean with $aE(X)+b$ and variance with $a^2Var(X)$. Keep $a$ in brackets when negative to avoid losing the square in variance.

Diagram of scaling and shifting

Verification checks

• Range reasonableness: The transformed mean should align with how outcomes were scaled and shifted, so a quick mental estimate should support the numeric result. If scaling is large, the mean and standard deviation should change comparably in magnitude.
• Variance sanity: Variance cannot be negative, and adding a constant must not change it. If your variance changed after pure translation, the setup is wrong.

Mean vs variance response

• Different sensitivities: Mean tracks location, so both scaling and shifting affect it. Variance tracks spread, so only scaling affects it and translation has no effect.
• Quick comparison table: | Feature | Mean $E(aX+b)$ | Variance $Var(aX+b)$ | | --- | --- | --- | | Effect of $a$ | Multiplied by $a$ | Multiplied by $a^2$ | | Effect of $b$ | Adds $b$ | No change | | Sign of $a$ | Can change sign of mean contribution | Never makes variance negative |

Reflection vs translation

• Negative scaling: If $a<0$, the distribution is reflected and stretched/compressed, which can move the mean across zero. Variance is unchanged by reflection direction because squaring removes sign.
• Pure shift case: When $a=1$, only a location change occurs, so shape and spread remain identical. This case is useful for fast checks in exam settings.

Fast exam workflow

• Identify $a$ and $b$ first: Rewrite the expression into exact $aX+b$ form before substituting values. This single step prevents most sign and coefficient errors.
• Compute in order: Find transformed mean first, then transformed variance, then standard deviation if needed as $\sqrt{Var}$. Keeping this order reduces mixing formulas under time pressure.

High-yield checks

• Memorize this anchor: > Always true: $Var(X+c)=Var(X)$ and $Var(cX)=c^2Var(X)$. These two identities recover the full rule immediately.
• Error-trap scan: After finishing, ask whether you accidentally used $a$ instead of $a^2$ in variance or added $b$ to variance. These are the two most common mark-losing mistakes.

Frequent mistakes

• Confusing $E(X^2)$ with $(E(X))^2$: These are generally different because averaging and squaring do not commute for random variables. Mixing them corrupts both variance and transformed-variance calculations.
• Dropping negative signs: In expressions like $2-X$, students often treat $a$ as $+1$ instead of $-1$. Correctly identifying $a=-1$ is essential even though $a^2=1$ in variance.

Misreading what variance measures

• Variance is not a location metric: It measures average squared distance from the mean, so adding a constant to every value does not change distances from the new mean. This is why translation does not alter variance.
• Standard deviation follow-up: If variance scales by $a^2$, standard deviation scales by $|a|$, not $a$. Forgetting absolute value can produce impossible negative standard deviations.