1、离散/连续随机变量、概率函数

  • 对于随机变量,均有概率,即有对应的分布特征。
  • Probability Distribution(概率分布)
    • Specifies the probabilities of the possible outcomes of a random variable.
  • Discrete/Continuous random variables(离散/连续随机变量)
    • Discrete random variables(离散随机变量)
      • take on at most a countable number of possible outcomes but do not necessarily to be limited.
      • 取值可数
    • Continuous random variables(连续随机变量)
      • cannot describe the possible outcomes of a continuous random variable(05)Common Probability Distribution - 图1with a list(05)Common Probability Distribution - 图2because the outcome(05)Common Probability Distribution - 图3not in the list, would always be possible.
      • 取值不可数
      • 特征:
        • (05)Common Probability Distribution - 图4(even though(05)Common Probability Distribution - 图5can happen)
        • (05)Common Probability Distribution - 图6(计算概率时通常是计算区间的概率)
  • Probability function(概率函数(05)Common Probability Distribution - 图7
    • For discrete random variables
    • (05)Common Probability Distribution - 图8
    • (05)Common Probability Distribution - 图9
  • Probability density function(PDF,概率密度函数(05)Common Probability Distribution - 图10
    • For continuous random variable commonly.
    • 用密度函数图形的面积表示概率(以下为正太分布的概率密度函数)

image.png

  • Cumulative probability function(CPF,累积概率函数(05)Common Probability Distribution - 图12
    • (05)Common Probability Distribution - 图13

image.png

2、概率分布:离散分布(2 种)

(1)Discrete uniform distribution(离散均匀分布)

  • A known, finite number of outcomes equally likely to happen.(取值确定、取值概率相等)
  • Every one of(05)Common Probability Distribution - 图15outcomes has equal probability(05)Common Probability Distribution - 图16
  • 如:Rolling a dice will have 6 possible outcomes as(05)Common Probability Distribution - 图17,the probability of each outcome is 0.167

    (2)Binomial distribution(二项式分布)

  • Bernoulli random variable(伯努利随机变量)

    • (05)Common Probability Distribution - 图18
    • (05)Common Probability Distribution - 图19
    • 一个事件只有两种结果的情况(如抛硬币,结果只能是正面 or 反面)
  • Binomial random variable(二项式随机变量)
    • the probability of(05)Common Probability Distribution - 图20successes in(05)Common Probability Distribution - 图21trails
    • (05)Common Probability Distribution - 图22
    • 如:计算抛(05)Common Probability Distribution - 图23次硬币,有(05)Common Probability Distribution - 图24次正面朝上(假设代表成功)的概率(每次实验都相互独立,且每次实验均只能有两种结果)
      • (05)Common Probability Distribution - 图25次实验
        • (05)Common Probability Distribution - 图26次成功,成功的概率用(05)Common Probability Distribution - 图27表示
        • (05)Common Probability Distribution - 图28次不成功
      • 公式原理:从(05)Common Probability Distribution - 图29次实验中挑选出(05)Common Probability Distribution - 图30次(即:(05)Common Probability Distribution - 图31),乘以其概率(05)Common Probability Distribution - 图32,剩下的(05)Common Probability Distribution - 图33次均为不成功的,其概率为(05)Common Probability Distribution - 图34,相乘即为所求概率。
    • 应用:可应用到预测生孩子判断性别概率上,如:医院里当天生(05)Common Probability Distribution - 图35个孩子,求生出(05)Common Probability Distribution - 图36个男孩的概率
  • 伯努利随机变量 vs 二项式随机变量
    • 相同点:事件均只有两种结果
    • 不同点:
      • 伯努利随机变量中,只做一次实验
      • 二项式随机变量中,做了 n 次实验
    • Expectations(期望)、variances(方差)【了解,不太考,推导见视频】

image.png

3、概率分布:连续分布(3 种)

(1)Continuous Uniform Distribution(连续均匀分布)

  • All intervals of the same length on the Continuous Uniform Distribution’s support are equally probable.
    • The support is defined by the two parameters, (05)Common Probability Distribution - 图38and(05)Common Probability Distribution - 图39, which are its minumum and maximum values.
  • 特征:

    • For all(05)Common Probability Distribution - 图40
      • (05)Common Probability Distribution - 图41
    • (05)Common Probability Distribution - 图42(不在可选范围内的结果的概率均为 0)

      (2)Normal distribution(正太分布)

      image.png
  • 特征:

    • (05)Common Probability Distribution - 图44(决定正太分布的变量:均值和方差)
    • Symmetrical distribution:skewness=0, kurtosis=3
      • 密度函数的图形形状对称
    • A linear combination of normally distributed random variables is also normally distributed.
      • 如果(05)Common Probability Distribution - 图45(05)Common Probability Distribution - 图46均服从正太分布,则他们的线性组合也符合正太分布
    • The tails get thin and go to zero but extend iinfinitely, asympotic(渐近)

      (a)Confidence intervals(置信区间,正太分布的置信区间)

  • 置信区间:个体落在某个区间的概率有多大(一个区间对应一个概率)

  • 68% confidence interval is(05)Common Probability Distribution - 图47
  • 90% confidence interval is(05)Common Probability Distribution - 图48
  • 95% confidence interval is(05)Common Probability Distribution - 图49
  • 99% confidence interval is(05)Common Probability Distribution - 图50

image.png

  • 之前讲过的切比雪夫不等式对于任何分布均成立,其含义为:

    • (05)Common Probability Distribution - 图52
    • 如,正太分布中,(05)Common Probability Distribution - 图53

      (b)标准正太分布(standard normal distribution)

  • 标准正太分布:(05)Common Probability Distribution - 图54or (05)Common Probability Distribution - 图55

  • standardization(普通正太分布的标准化):
    • 如果(05)Common Probability Distribution - 图56,则(05)Common Probability Distribution - 图57
    • 推导:
      • 如果(05)Common Probability Distribution - 图58,则:
        • (05)Common Probability Distribution - 图59
        • (05)Common Probability Distribution - 图60
        • (05)Common Probability Distribution - 图61
      • 如果(05)Common Probability Distribution - 图62,则:
        • (05)Common Probability Distribution - 图63
        • (05)Common Probability Distribution - 图64 (一组数据同时加上一个数,不会影响其离散程度)
        • (05)Common Probability Distribution - 图65
  • 为什么要进行标准化?
    • 方便查表(Z-table 表);统计学家不可能做无数个正太分布表。
      • (05)Common Probability Distribution - 图66
      • (05)Common Probability Distribution - 图67
      • (05)Common Probability Distribution - 图68
    • Z-table 表如下(每个数值均为累积概率,只给出了(05)Common Probability Distribution - 图69的部分,另外一部分对称):

image.png

  • 常见考法:

    • 给定(05)Common Probability Distribution - 图71,求(05)Common Probability Distribution - 图72
      • 先进行标准化:
        • (05)Common Probability Distribution - 图73
        • (05)Common Probability Distribution - 图74
    • 计算(05)Common Probability Distribution - 图75
      • 画图,基于对称原理,可得:(05)Common Probability Distribution - 图76
    • 计算(05)Common Probability Distribution - 图77
      • 画图,基于对称原理,可得:
        • (05)Common Probability Distribution - 图78

          (c)名词概念:Shortfall risk、Roy’s safety-first criterion、Maximize S-F-Ratio

  • Shortfall risk

    • (05)Common Probability Distribution - 图79:threshold level return、minimum return required
    • 如果(05)Common Probability Distribution - 图80,出现这种情况的概率((05)Common Probability Distribution - 图81),即为 Shortfall risk
    • 即:基金经理做投资的回报率小于客户最底要求回报率的风险
    • (05)Common Probability Distribution - 图82越小越好
  • Roy’s safety-first criterion
    • (05)Common Probability Distribution - 图83
    • 类似夏普比率,但此处的基准是(05)Common Probability Distribution - 图84(客户要求的最低回报率),而夏普比率的基准为(05)Common Probability Distribution - 图85(无风险回报率)
    • (05)Common Probability Distribution - 图86越大越好
  • Maximize S-F-Ratio

    • 如果(05)Common Probability Distribution - 图87越大,则(05)Common Probability Distribution - 图88越小

      (3)lognormal distribution(对数正太分布)

  • If(05)Common Probability Distribution - 图89is normal, then(05)Common Probability Distribution - 图90is lognormal.

    • 如果(05)Common Probability Distribution - 图91是正太分布,则(05)Common Probability Distribution - 图92服从对数正太分布(注意:不能颠倒说法)
  • 特点:Right skewed(右偏).
    • 考试时,可能会与偏度的相关概念与特性结合起来考
  • The values of random variables that follow lognormal distribution are always be positive, so it is useful for modeling asset prices.
    • 资产价格(P)总是大于 0 的,因此实际应用中总是假设资产价格符合对数正太分布。
    • 而 Return 可以小于 0 或大于 0,与正太分布的特性相似,故实际应用中假设 Return 是符合正太分布的。

image.png

4、Monte Carlo simulation、Historical simulation

  • 离散(Discrete)
    • (05)Common Probability Distribution - 图94
  • 连续(Continuous)
    • (05)Common Probability Distribution - 图95
  • (05)Common Probability Distribution - 图96(持有一年)
  • (05)Common Probability Distribution - 图97(持有 T 年)
  • Monte Carlo simulation
    • generate a large number of random samples from specified probability distribution(s) to represent the operation of risk in the system.
    • It is used in planning, in financial risk management, and in valuing complex securities.
    • Limitations:
      • The operating of Monte Carlo simulation is very complex and we must assume a parameter distribution in advance.
      • Monte Carlo simulation provides only statistical estimates, not exact results.
  • Historical simulation

    • repeat sampling from a historical data series.
    • grounded in actual data but can reflect only risks represented in the sample historical data.
    • Compared with Monte Carlo simulation, historical simulation does not lend itself to “what if” analyses.

      5、例题

      (1)离散/连续随机变量、概率函数、概率分布

      (a)离散型随机变量的概率分布

      image.png

      (b)连续均匀分布、累积分布函数

      image.png

      (c)概率分布的理解:离散均匀分布

      image.png

      (d)二项式分布求概率公式

      image.png

      (e)二项式分布:求期望 E(X)

      image.png

      (f)二项式分布:概念理解

      image.png

      (2)连续分布

      (a)连续均匀分布

      image.png

      (b)正太分布:Confidence intervals

      image.png

      (c)正太分布:基于区间范围计算均值和标准差

      image.png

      (d)标准正太分布:概率计算

      image.png

      (e)Shortfall risk、Roy’s safety-first criterion、Maximize S-F-Ratio

      Roy’s safety-first criterion 计算与比较

      image.png
  • (05)Common Probability Distribution - 图109

  • (05)Common Probability Distribution - 图110 (求解取最大值即可)

    Shortfall risk 的理解、Roy’s safety-first criterion 计算与比较、正太分布概率计算

    image.png
    image.png

    (f)对数正太分布

    image.png

    (3)Monte Carlo simulation

    image.png