Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- \documentclass[12pt]{article}
- \usepackage[utf8]{inputenc}
- \usepackage{amsmath}
- \usepackage{geometry}
- \geometry{
- a4paper,
- total={170mm,257mm},
- left=20mm,
- top=20mm,
- }
- \pagenumbering{gobble}
- \begin{document}
- Distributions of settlement populations are modeled using Pareto distributions, probability density function: \\
- $ f(x) = \frac{\alpha x^{\alpha}_m}{x^{\alpha + 1}} $ \\
- cumulative density function: \\
- $ F(x) = 1 - (\frac{x_m}{x})^\alpha $ \\
- expected value: \\
- $E = \frac{\alpha x_m}{\alpha - 1}$ \\
- where $x_m$ – threshold value, $\alpha$ – shape parameter.
- A set of values of population sizes can be expressed as a (sorted) series of realizations of identical independent Pareto distributions: \\
- $X = (X_{1:n}, ..., X_{n:n}), \quad i \leq j \rightarrow X_{i:n} \leq X_{j:n}, \quad X_{i:n} \sim P(x_m, \alpha)\forall i \in [n]$ \\
- where $n$ – number of observations. The probability density of the biggest value in set ($X_{n:n}$) being $m$: \\
- \begin{align*}
- &\mathcal{P} = P[X_{n:n} = m\ \wedge \bigwedge_{i=1}^{n-1}X_{i:n}\leq m] \overset{independence}{=} \\ &P[X_{n:n} = m] * P[\bigwedge_{i=1}^{n-1}X_{i:n}\leq m] \overset{independence}{=} \\
- &P[X_{n:n} = m] * \prod_{i=1}^{n-1}P[X_{i:n}\leq m] \overset{sameness}{=} \\
- &P[X_{n:n} = m] * \prod_{i=1}^{n-1}P[X_{1:n}\leq m] = \\
- &P[X_{n:n} = m] * (P[X_{1:n}\leq m])^{n-1} = \\
- &f(X_{n:n})*(F(X_{n:n}))^{n-1} = \\
- &\frac{\alpha x^{\alpha}_m}{X_{n:n}^{\alpha + 1}}*(1 - (\frac{x_m}{X_{n:n}})^\alpha)^{n-1}
- \end{align*}
- Putting $\gamma = \frac{x_m}{X_{n:n}}$:
- \begin{align*}
- \mathcal{P}(\alpha) = \frac{1}{X_{n:n}}(\alpha \gamma^\alpha(1-\gamma^\alpha)^{n-1})
- \end{align*}
- Maximum likelihood estimator:
- \begin{align*}
- &\hat{\alpha} = \underset{\alpha}{maxarg}\{\mathcal{P}(\alpha)\} = \underset{\alpha}{maxarg}\{ln(\mathcal{P}(\alpha))\} = (ln(\mathcal{P}(\alpha)))'_0 = \\ &(ln(\alpha) +\alpha ln(\gamma) + (n-1)ln(1-\gamma^\alpha))'_0 = \\
- &\left(\frac{1}{\alpha}+ln(\gamma)-(n-1)\frac{1}{1-\gamma^\alpha} ln(\gamma)\gamma^\alpha\right)_0
- \end{align*}
- The expected overall population is the number of colonies above the threshold population times the expected population of a colony above the threshold divided by the percentage of people who live in those: \\
- $\hat{N} = n\hat{E}:p = n\frac{\hat{\alpha} x_m}{\hat{\alpha} - 1}:p$
- \end{document}
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement