Supplemental Proofs

Other vignettes in this package provide extensive derivations of the algorithms formulas used to calculate various aspects of the study design. In some cases, these derivations require additional proofs that are not directly related to the topic. In order to avoid tangents, these supplemental proofs are gathered here for reference.

Table of Contents

  1. Expression of Unequal Sample Sizes as a Proportion
  2. Expression of Unequal Sample Sizes as a Ratio

1. Expression of Unequal Sample Sizes as a Proportion

Let $X_1$ represent a sample of size $n_1$ from a population.
Let $X_2$ represent a sample of size $n_2$ from a population.
Define $n = n_1 + n_2$ where $n_1 \leq n$ and $n_2 \leq n$.

Since $n_2 \leq n$, we may write $n_2 = k \cdot n$ where $0 \leq k \leq 1$.
Additionally, $$n = n_1 + n_2$$ $$\Rightarrow n_1 = n - n_2$$

By substitution $$n_1 = n - n_2$$ $$n_1 = n - k \cdot n$$ $$n_1 = n \cdot (1 - k)$$ $$n_1 = (1 - k) \cdot n$$

Observing that $n_1 = n \cdot (1-k)$ and $n_2 = k \cdot n$, we see that both $n_1$ and $n_2$ may be expressed strictly in terms of $n$ whenever $n$ is the sum of $n_1$ and $n_2$.

1.a Practical Implementation

Being able to express $n_2$ in terms of $n_1$ makes it easier to calculate sample size and power calculations in the case of unequal sample sizes. In this particular context, we can express $n_2$ in terms of a proportion of $n$.
For example, if there are two subjects in Group 1 for every subject in Group 2, $n_2 = \frac{1}{3} \cdot n$ and $n_1 = \frac{2}{3} \cdot n$.

2. Expression of Unequal Sample Sizes as a Ratio

Let $X_1$ represent a sample of size $n_1$ from a population.
Let $X_2$ represent a sample of size $n_2$ from a population.
Define $n = n_1 + n_2$ where $n_1 \leq n$ and $n_2 \leq n$.
Since $n_2 \leq n$, we may write $n_2 = k \cdot n$ where $0 \leq k \leq 1$.

It is demonstrated in Section 1 that if $n_2 = k \cdot n$ then $n_1 = (1 - k) \cdot n$. It follows that $$k = \frac{n_2}{n}$$ and $$(1-k) = \frac{n_1}{n}$$

Now we observe that $$n = n_1 + n_2 $$ $$\Rightarrow n = n_1 + \frac{n_2 \cdot n_1}{n_1} $$ $$\Rightarrow n = n_1 + \frac{n_2}{n_1} \cdot n_1 $$ $$\Rightarrow n = n_1 + \frac{\frac{n_2}{n}}{\frac{n_1}{n}} \cdot n_1 $$ $$\Rightarrow n = n_1 + \frac{k}{1-k} \cdot n_1 $$ $$\Rightarrow n_1 + n_2 = n_1 + \frac{k}{1-k} \cdot n_1 $$ $$\Rightarrow n_2 = n_1 + \frac{k}{1-k} \cdot n_1 - n_1 $$ $$\Rightarrow n_2 = n_1 - n_1 + \frac{k}{1-k} \cdot n_1 $$ $$\Rightarrow n_2 = \frac{k}{1-k} \cdot n_1 $$

Thus, when $n = n_1 + n_2$ we may express $n_2$ in terms of $n_1$.

2.a Practical Implementation

Being able to express $n_2$ in terms of $n_1$ makes it easier to calculate sample size and power calculations in the case of unequal sample sizes. In this particular context, we can express $n_2$ in terms of a ratio multiplied by $n_1$.
For example, if there are two subjects in Group 1 for every subject in Group 2, $n_2 = \frac{1}{2} \cdot n_1$.



nutterb/junkyard documentation built on May 24, 2019, 10:51 a.m.