Sum of Squares (SSx) , often written as , is a key value used to measure the total variation of a single variable (
). It is a foundational step for calculating variance, standard deviation, and the slope in linear regression.
In simple terms, Sxx tells you how much your data points "spread out" from their own average. The Formulas
There are two ways to calculate it. Both give the same result, but one is usually easier for hand calculations. 1. The Definitional Formula
Use this to understand the logic: subtract the mean from each point, square the result, and add them all up.
cap S x x equals sum of open paren x sub i minus x bar close paren squared 2. The Computational Formula
Use this for faster math or when working with large datasets: Sxx Variance Formula
cap S x x equals sum of x squared minus the fraction with numerator open paren sum of x close paren squared and denominator n end-fraction sum of x squared Square every number first, then add them up. Add all the numbers first, then square the total. The total number of data points. Why is it useful? Sxx is the "numerator" for variance. If you want the actual Variance ( , you just divide Sxx by the degrees of freedom:
s squared equals the fraction with numerator cap S x x and denominator n minus 1 end-fraction A Quick Example If your data is correlation coefficient
Sample Variance ( formula—often denoted as cap S sub x x end-sub
in the context of sum of squares—measures how much a set of numbers spreads out from their average. In simple terms, cap S sub x x end-sub represents the Sum of Squared Deviations
from the mean. Here is the breakdown of how to understand and calculate it. 1. The Formula
There are two ways to write this. The "definitional" version helps you understand the logic, while the "computational" version is much faster for manual math. The Definitional Formula Sum of Squares (SSx) , often written as
cap S sub x x end-sub equals sum of open paren x sub i minus x bar close paren squared : Each individual value in your data set. : The mean (average) of the data. : The sum of all those squared differences. The Computational (Shortcut) Formula This is usually easier if you are using a calculator:
cap S sub x x end-sub equals sum of x squared minus the fraction with numerator open paren sum of x close paren squared and denominator n end-fraction 2. Step-by-Step Calculation If you have a small data set, like , here is how you find cap S sub x x end-sub using the definitional method: Find the Mean ( Subtract Mean from each point: Square those results: Sum them up ( cap S sub x x end-sub cap S sub x x end-sub vs. Sample Variance ( It is important to note that cap S sub x x end-sub is not the final variance . It is the numerator used to find it. To get the Sample Variance ( , you divide cap S sub x x end-sub To get the Population Variance ( sigma squared , you divide cap S sub x x end-sub In our example above ( Sample Variance: 4. Why "Squared"?
We square the differences because if we just added them up ( ), they would equal
. Squaring ensures all values are positive, giving us a meaningful "total distance" from the center. 5. Common Use Cases Linear Regression: cap S sub x x end-sub is a foundational piece for calculating the slope ( ) of a regression line. Standard Deviation:
Once you have the variance, you take the square root to find the standard deviation. is used to calculate the slope of a regression line
Sxx is formally defined as the sum of squared deviations of each data point from the mean. It is a measure of total variability in the independent variable (x). Dividing Sxx by (n-1) yields the sample variance: Definition and Interpretation Sxx is formally defined as
[ s_x^2 = \fracS_xxn-1 = \frac\sum (x_i - \barx)^2n-1 ]
Thus, Sxx is the numerator of the variance formula. It captures the raw dispersion before scaling by degrees of freedom. A larger Sxx indicates greater spread of (x) values.
σ^2 = Sxx / n
s^2 = Sxx / (n − 1)
Let’s start with the most common definition. Given a set of ( n ) observations for a variable ( x ): ( x_1, x_2, x_3, \dots, x_n ), the quantity Sxx is defined as:
[ S_xx = \sum_i=1^n (x_i - \barx)^2 ]
Where:
This is often called the corrected sum of squares (or sum of squares about the mean). It measures the total squared deviation of each data point from the average.
If we simply summed ( (x_i - \barx) ), the result would always be zero (positive and negative deviations cancel). Squaring removes the sign, ensuring we measure magnitude of spread, not direction.