Identifying T, Chi-Square, and F Variables from Formulas

Reading a Formula as a Distributional Construction

A formula can reveal the distribution of a random variable, but only when its ingredients and their relationships are known. In particular, the same-looking expression may have a different distribution if its normal variables are not standard, if required independence fails, or if a scaling factor is missing. The useful habit is to identify the random ingredients first, then compare the formula with the defining construction of a t, chi-square, or F variable.

This tutorial focuses on recognition from formulas. It does not require calculating tail probabilities or quantiles. Those calculations come after the distribution and its degrees of freedom have been identified.

Definition: A standard normal random variable has distribution \(Z\sim N(0,1)\). If \(Z_1,\ldots,Z_k\) are independent standard normal random variables, then their sum of squares \(Q=Z_1^2+\cdots+Z_k^2\) has a chi-square distribution with \(k\) degrees of freedom, written \(Q\sim\chi^2_k\). If \(Z\sim N(0,1)\) is independent of \(Q\sim\chi^2_\nu\), then \(T=Z/\sqrt{Q/\nu}\) has a t distribution with \(\nu\) degrees of freedom, written \(T\sim t_\nu\). If \(Q_1\sim\chi^2_{d_1}\) and \(Q_2\sim\chi^2_{d_2}\) are independent, then \(F=(Q_1/d_1)/(Q_2/d_2)\) has an F distribution with numerator degrees of freedom \(d_1\) and denominator degrees of freedom \(d_2\), written \(F\sim F_{d_1,d_2}\).

These constructions tell us what to look for. A chi-square formula is a sum of squared independent standard normals. A t formula has one standard normal in its numerator and the square root of an independent chi-square variable divided by its degrees of freedom in its denominator. An F formula is a ratio of two independent chi-square variables, each divided by its own degrees of freedom.

Identify the Ingredients Before Naming the Distribution

For formulas built literally from finite sums of squared standard normals, the degrees of freedom are the number of independent squared terms. For example, \(Z_1^2+Z_2^2+Z_3^2+Z_4^2\) has four degrees of freedom when the four \(Z_i\) are independent standard normals. The variables must be standard normal: squaring a normal variable with a nonzero mean or a variance other than one does not generally give the stated chi-square construction.

The chi-square, t, and F distribution families are also defined for positive real degrees of freedom, not just integers. Thus an expression made from a finite count of squared standard normals directly supplies an integer degrees of freedom, while a distribution may still be specified with a positive noninteger degree of freedom. When reading a formula, distinguish the degrees of freedom given by its explicit finite sum from the broader range allowed by the distribution family.

Independence is another essential part of the pattern. The normal in a t numerator must be independent of the chi-square quantity in its denominator. The two chi-square variables in an F ratio must be independent. Also, the scaling must match: the F construction divides each chi-square variable by its own degrees of freedom. A ratio with an extra constant multiplier is generally a scaled F variable, not an F variable with the same degrees of freedom.

Worked Example: Recognizing a Chi-Square Variable

Suppose \(Z_1,\ldots,Z_6\) are independent, and each has distribution \(N(0,1)\). Consider \(Q=Z_1^2+Z_2^2+Z_3^2+Z_4^2+Z_5^2+Z_6^2\). Identify its distribution and degrees of freedom.

Each of the six terms is the square of a standard normal variable, and the six variables are independent. The formula is exactly the defining sum-of-squares construction, so

Q=\sum_{i=1}^{6}Z_i^2\sim\chi^2_6.

There are six squared terms, so the degrees of freedom are \(6\). If the formula instead contained \(Z_1^2+\cdots+Z_6^2\) with dependent \(Z_i\), the count of terms alone would not establish a \(\chi^2_6\) distribution.

Recognizing a t Variable

To check whether a formula has a t distribution, locate the numerator and denominator separately. The numerator must be standard normal. The denominator must be the square root of a chi-square variable divided by its degrees of freedom, and that chi-square variable must be independent of the numerator. The degrees of freedom for the t distribution are the degrees of freedom of that chi-square variable.

Worked Example: A t Variable Built From Eight Squares

Suppose \(Z,Y_1,\ldots,Y_8\) are mutually independent standard normal random variables. Define \(T=Z/\sqrt{(Y_1^2+\cdots+Y_8^2)/8}\). Identify the distribution.

The denominator's sum has eight independent squared standard normal terms. Therefore \(Q=Y_1^2+\cdots+Y_8^2\sim\chi^2_8\). The numerator \(Z\) is standard normal and independent of \(Q\), and the denominator is \(\sqrt{Q/8}\). Substitution into the defining t construction gives

T=\frac{Z}{\sqrt{Q/8}}\sim t_8.

The t degrees of freedom are \(8\), not \(9\): they come from the chi-square sum in the denominator. No degrees of freedom are added for the separate numerator \(Z\).

The square root and division by the degrees of freedom are both important. A formula \(Z/\sqrt{Q}\) with \(Q\sim\chi^2_\nu\) does not match the standard t construction unless the denominator is rewritten as \(\sqrt{Q/\nu}\) with the corresponding adjustment to the whole expression. Likewise, \((Z/\sqrt{Q/\nu})^2\) is not itself a t variable: it is the square of a t variable.

Recognizing an F Variable

An F construction has two chi-square pieces. First identify the numerator sum of squares and its degrees of freedom; then identify the denominator sum and its degrees of freedom. Divide each chi-square variable by its own degrees of freedom before taking their ratio. The order matters for the labels: the first degrees of freedom are the numerator degrees of freedom, and the second are the denominator degrees of freedom.

Worked Example: Reading Both Degrees of Freedom in an F Ratio

Suppose \(A_1,\ldots,A_5,B_1,\ldots,B_9\) are mutually independent standard normal random variables. Define

R=\frac{(A_1^2+\cdots+A_5^2)/5}{(B_1^2+\cdots+B_9^2)/9}.

The numerator sum \(U=A_1^2+\cdots+A_5^2\) is \(\chi^2_5\), and the denominator sum \(V=B_1^2+\cdots+B_9^2\) is \(\chi^2_9\). The groups of normal variables are independent, so \(U\) and \(V\) are independent. The expression divides \(U\) by \(5\), divides \(V\) by \(9\), and takes their ratio. Hence

R=\frac{U/5}{V/9}\sim F_{5,9}.

The numerator degrees of freedom are \(5\), and the denominator degrees of freedom are \(9\). Reversing the ratio would switch the order and give an \(F_{9,5}\) variable, not \(F_{5,9}\).

A Short Justification of the Recognition Rules

The rules above are applications of the distribution definitions, rather than visual similarities alone. The chi-square rule follows directly by setting \(Q=\sum_{i=1}^{k}Z_i^2\) for \(k\) independent standard normals; by definition, \(Q\sim\chi^2_k\). Once that identification is made, the t and F rules follow by matching the remaining algebra to their defining ratios.

For a t formula, suppose \(Q=\sum_{i=1}^{\nu}Y_i^2\), where the \(Y_i\) are independent standard normals and also independent of \(Z\sim N(0,1)\). Then \(Q\sim\chi^2_\nu\), and the expression becomes

\frac{Z}{\sqrt{(\sum_{i=1}^{\nu}Y_i^2)/\nu}} = \frac{Z}{\sqrt{Q/\nu}} \sim t_\nu.

For an F formula, let \(U\sim\chi^2_{d_1}\) and \(V\sim\chi^2_{d_2}\) be independent. The expression \((U/d_1)/(V/d_2)\) matches the definition of \(F_{d_1,d_2}\), so it has that distribution. These arguments also show why checking the ingredients and the scaling is necessary: without them, the formula does not match the stated construction.

In these sum-of-squares constructions, the degrees of freedom are positive integers because each chi-square variable is a sum of that many independent squared standard normals. The chi-square, t, and F distribution families are also defined for positive real degrees of freedom. A formula explicitly showing a finite number of squared terms therefore gives an integer count, while a distribution specified directly by a positive real degree of freedom need not arise from that particular finite-sum representation.

Expressions That Do Not Match Exactly

A formula should not be assigned a familiar distribution merely because it contains squares, a normal variable, or a ratio. Check the standardization, independence, and scaling conditions explicitly. For example, if \(Z\sim N(0,1)\) and \(Q\sim\chi^2_4\) are independent, then \(Z/\sqrt{Q/4}\sim t_4\). But

\frac{Z}{\sqrt{Q}}=\frac{1}{2}\frac{Z}{\sqrt{Q/4}}.

This is one-half of a \(t_4\) variable, so it is a scaled t variable rather than a standard \(t_4\) variable. The difference in the denominator changes the distribution.

Worked Example: Checking Scaling in an F-Like Ratio

Suppose \(U\sim\chi^2_3\) and \(V\sim\chi^2_7\) are independent, and consider \(W=(U/3)/(2V/7)\). Is \(W\) an \(F_{3,7}\) variable?

The standard F ratio would be \((U/3)/(V/7)\). Here the denominator is twice as large, so

W=\frac{U/3}{2V/7} =\frac{1}{2}\frac{U/3}{V/7}.

The ratio \((U/3)/(V/7)\) has distribution \(F_{3,7}\); therefore \(W\) is one-half of an \(F_{3,7}\) variable, not an \(F_{3,7}\) variable itself. The degrees of freedom identify the unscaled ratio, but they do not absorb an extra multiplier.

A Reliable Classification Routine

When given an unfamiliar expression, rewrite it into named components before deciding what distribution it has. This makes missing assumptions and misplaced degrees of freedom easier to spot.

Mark the ingredients.
Identify every normal variable, squared term, sum, and ratio. Confirm that any claimed standard normal has mean \(0\) and variance \(1\).

Check independence.
Check independence among the normal terms in each sum and between components that will form a t or F ratio.

Match the structure and scaling.
Match a sum of squares to \(\chi^2_k\), a standard normal divided by \(\sqrt{Q/\nu}\) to \(t_\nu\), or a ratio \((U/d_1)/(V/d_2)\) to \(F_{d_1,d_2}\).

State the conclusion with its conditions.
Name the distribution and degrees of freedom, and mention any required independence or scaling condition that is not already explicit.

Common Mistakes and Why the Distinctions Matter

Correct classification matters because probabilities and quantiles depend on the exact distribution and degrees of freedom. Choosing the wrong degrees of freedom, reversing numerator and denominator, or overlooking a multiplier can produce an incorrect tail area even if the formula looks familiar.

Counting every symbol instead of squared terms. In a chi-square sum, count the independent squared standard normal variables, not the number of letters appearing elsewhere in the expression.
Ignoring independence. A formula with the right-looking pieces is not enough if the normal and chi-square components in a t ratio, or the two chi-square components in an F ratio, are dependent.
Forgetting the scaling. The t denominator is \(\sqrt{Q/\nu}\), and an F ratio divides each chi-square component by its own degrees of freedom.
Swapping F labels. In \(F_{d_1,d_2}\), \(d_1\) belongs to the numerator and \(d_2\) to the denominator.
Assuming all distribution degrees of freedom must be integer counts. Explicit finite sums give integer counts, but the chi-square, t, and F families are defined for any positive real degrees of freedom.

Key takeaway: Classify the pieces first and the whole expression second. Standardization, independence, and the exact scaling determine whether a formula is chi-square, t, F, or merely a scaled version of one of these variables.

Check Your Understanding

For each question, identify the distribution if the stated conditions are sufficient, and explain which feature of the formula determines the degrees of freedom.

Let \(X_1,\ldots,X_4\) be independent standard normal variables. What is the distribution of \(X_1^2+\cdots+X_4^2\)?
Let \(Z\sim N(0,1)\) be independent of \(Q\sim\chi^2_{11}\). What is the distribution of \(Z/\sqrt{Q/11}\)?
Let \(U\sim\chi^2_6\) and \(V\sim\chi^2_2\) be independent. Identify the distribution of \((U/6)/(V/2)\), including the order of the degrees of freedom.
Why does a ratio \((U/6)/(3V/2)\), with the \(U\) and \(V\) in question 3, not have the same distribution as the standard F ratio?
Does the fact that explicit sums of squared standard normals have integer degrees of freedom mean that the t, chi-square, and F families only permit integer degrees of freedom? Explain.