# 统计代写|贝叶斯统计代写Bayesian statistics代考|STAT206

## 统计代写|贝叶斯统计代写Bayesian statistics代考|EXAMPLES: MULTINOMIAL MODELS

Categorized data, such as one encounters frequently in genetics (where the categories are usually phenolypes), give rise to multinomial models. Since the distribution theory of multinomial models is well known and especially elegant, they will serve the illustrative purposes of this section well. Some slight familiarity with Mendelian genetics will be assumed. ${ }^4$

EXAMPLE 1 (genetic linkage). Two factors are ‘linked’ when they lie on the same chromosome. As is customary, we denote the dominant alleles by capital letters, and reserve small letters for the recessive alleles. The heterozygote can occur in either of two phases: coupling $(A B a b)$ where $A$ and $B$ lie on one chromosome of a homologous pair and $a, b$ on the other, or repulsion $(A b a B)$, where a dominant $A$ is linked to a recessive $b$ and vice versa. Chromosomes may break and members of a pair sometimes exchange homologous sections. As a result, the two chromosomes $A b$ and $a B$ of repulsion phase may give rise to the pair $A B, a b$ of coupling phase (or vice versa). The rejoined chromosomes are called recombinants and crossing-over is said to have occurred.

Let $\theta$ be the probability of crossing-over (called the ‘recombination fraction’). Then the gametic output of the heterozygote $A b a B$ will be $A B$, $A b, a B$, and $a b$ in the relative frequencies:
$$\text { (5.4) } \quad \theta / 2,(1-\theta) / 2,(1-\theta) / 2, \theta / 2 \text {. }$$
For crossing-over occurs with probability $\theta$, and then one member of each chromosome pair is $A B$ and the other is $a b$, and similarly for the non-recombinant types $A b$ and $a B$. If we cross the heterozygote to a double recessive, $a b a b$, a so-called ‘double backcross’, the four phenotypes $A b, A b$, $a B$, and $a b$ will also be genotypes, since the recessive genes from the doubly recessive parent have nil effect. Hence, (5.4) gives the category probabilities for the double backcross, the categories being the phenotypes (and genotypes) $A B, A b, a B, a b$. Let $n$ be the sample size, and let $n_i$ be the observed number of off-spring in category $i, i=1,2,3,4$.

## 统计代写|贝叶斯统计代写Bayesian statistics代考|MULTINOMIAL MODELS WITHOUT FREE PARAMETERS

Let the random vector $\left(X_1, \ldots, X_k\right)$ have the multinomial distribution with $k$ categories and category probabilities $p_1, \ldots, p_k$ :
$$P\left(X_1=x_1, \ldots, X_k=x_k\right)=\frac{n !}{x_{1} ! \ldots x_{k-1} ! x_{k} !} p_1^{x_1} \ldots p_k^{x_k}$$
where $x_i$ is the observed number in cateogry $i$ and $\Sigma x_i=n$, the sample size.

Then the vector $\left(X_1, \ldots, X_{k-1}\right)$ has, asymptotically, a non-singular normal distribution with density
$(5.12)(2 \pi)^{-(k-1) / 2} \Lambda^{-1 / 2} e^{-Q / 2}$,
where
$$Q=\sum_{i, j}^{k-1}\left(\Lambda_{i j} / \Lambda\right)\left(X_i-n p_i\right)\left(X_j-n p_j\right)$$
is the quadratic form whose matrix of coefficients is inverse to the covariance matrix of the multinomial with diagonal entries $n p_i\left(1-p_i\right)$ and off-diagonal entries $-n p_i p_j, i \neq j$. The $\Lambda_{i j}$ are thus the co-factors of the covariance matrix and $\Lambda$ the determinant. Given the obvious analogy to the univariate case, $\Lambda$ is called the generalized variance of the distribution and its square root the generalized standard deviation. It is well known that $Q$ has a chi square distribution with $k-1$ degrees of freedom. Hence
(5.13) $Q=\chi_{\alpha, k-1}^2$
is the equation of a $100(1-q) \%$ direct confidence ellipsoid for the vector $\left(X_1, \ldots, X_{k-1}\right)$, writing $\chi_{\alpha, k-1}^2$ for the upper $100 \alpha \%$ point of the chi square distribution with $k-1 d f$. (More generally, the experimental outcome will fall in a $100(1-\alpha) \%$ direct confidence region (or $D C R$ ) with probability $100(1-\alpha)$, it being assumed that the probability of every included outcome, conditional on the model, exceeds that of any excluded outcome.)

# 贝叶斯统计代考

## 统计代写|贝叶斯统计代写Bayesian statistics代考|EXAMPLES: MULTINOMIAL MODELS

(5.4) $\theta / 2,(1-\theta) / 2,(1-\theta) / 2, \theta / 2$.

## 统计代写|贝叶斯统计代写Bayesian statistics代考|MULTINOMIAL MODELS WITHOUT FREE PARAMETERS

$$P\left(X_1=x_1, \ldots, X_k=x_k\right)=\frac{n !}{x_{1} ! \ldots x_{k-1} ! x_{k} !} p_1^{x_1} \ldots p_k^{x_k}$$

$$(5.12)(2 \pi)^{-(k-1) / 2} \Lambda^{-1 / 2} e^{-Q / 2} \text { ， }$$

$$Q=\sum_{i, j}^{k-1}\left(\Lambda_{i j} / \Lambda\right)\left(X_i-n p_i\right)\left(X_j-n p_j\right)$$

myassignments-help数学代考价格说明

1、客户需提供物理代考的网址，相关账户，以及课程名称，Textbook等相关资料~客服会根据作业数量和持续时间给您定价~使收费透明，让您清楚的知道您的钱花在什么地方。

2、数学代写一般每篇报价约为600—1000rmb，费用根据持续时间、周作业量、成绩要求有所浮动(持续时间越长约便宜、周作业量越多约贵、成绩要求越高越贵)，报价后价格觉得合适，可以先付一周的款，我们帮你试做，满意后再继续，遇到Fail全额退款。

3、myassignments-help公司所有MATH作业代写服务支持付半款，全款，周付款，周付款一方面方便大家查阅自己的分数，一方面也方便大家资金周转，注意:每周固定周一时先预付下周的定金，不付定金不予继续做。物理代写一次性付清打9.5折。

Math作业代写、数学代写常见问题

myassignments-help擅长领域包含但不是全部: