FRM一级知识点框架
2020-03-17 15:01:48 1 举报
AI智能生成
FRM一级知识点框架
作者其他创作
大纲/内容
quantitative analysis
fundamental of probability
probability density function
cumulative distribution function
basic statistics
moment
一阶矩mean
衡量中心趋势
population mean(μ)
sample mean(X)
mean,median,mode
二阶矩variance
衡量数据dispersion程度
variance
sample variance考虑自由度
自由度:自由变化的变量个数
covariance
比正负,不能比大小(带单位),因此变换产生correlation coefficient去掉单位来进行比较
correlation coefficient
衡量linear relationship
取值-1到1
等于0只能证明没有linear 关系,但是不能说明数据independent
三阶矩skewness
衡量数据对称性
negative skewness
左偏,mean<medium<mode
positive skewness
右偏,mean>medium>mode
四阶矩kurtosis
衡量数据分布图尾巴薄厚层度
lept
尖峰肥尾,极端值概率增加
plat
矮峰瘦尾,极端值概率较小
只能比较SD相同分布的kurtosis
BLUE
衡量估计量好坏的标准
chebyshev' inequality
对于任何分布,P(X<k*SD)>1-(1/k^2)
distribution
binominal
mean=np
variance=np(1-p)
P如何计算
poisson distribution
mean=np
variance=np
P如何计算
continues uniform
mean=a+b/2
子主题
normal distribution
如何标准化?
lognormal distribution
描述资产价格
chi-square distribution
定义,取值范围,positive skewness
t-distribution
和normal distribution几乎一样,只是峰度不固定
主要用于描述N<30,variance unknown的normal population
由于mean已知,因此degree of freedom 为n-1
f distribution
右偏
hypothesis test
term
sample statistic
population parameter
standard error
SD of the sample means
Standard error=population/根号n
central limited thereom
将未知分布变换为正态分布
mean/variance/使用条件n大于30
point estimate/interval estimate
公式:X+k*standard error
Z和T检验的使用条件:n大于30用Z,总体方差已知用Z
steps of hypothesis test
设定H0和H1,希望拒绝的放在H0
计算检验统计量
mean
Z或者T
variance
chi-square或者F
X-μ/σ*n^-2
查表得出的数在拒绝域与否
得出结论,拒绝原假设与否
P—value
概率与概率进行比较,越小越拒绝
Type I和Type Ⅱ
confidence level和power of test
Type Ⅰ去真/Type 存伪
同时下降增加sample population
linear regression
sigle
TSS、ESS、SSR的关系
sqrR measure 解释力度=ESS/TSS
multi
adjusted sqrt R=1-{(n-1)/(n-k-1)*(1-sqrt R)}
multicollinear
如何检测
common sense
相关系数大于0.7
f test通过,t test通不过
hypothesis test of regression coefficient
如果k个自变量,回归自由度为n-k-1
residual variance保持不变就是homoskedastic
ANOVA table重点掌握
time series
trend
如何estimate trend
OLS
判断model好坏
MSE
sqrt S
AIC
SIC
cycle
dummy variable
noise
能被forcast前提,协方差平稳covariance stationary
mean和variance stable over time
autoregression和partial autoregression的区别
white noise
mean=0;variance is constant的noise
如果相互independent,那就是strong white noise
WOLD 理论
一个协方差稳定的process能被分解为有限个white noise process
modeling
MA
AR
ARMA
estimate volatility
收益率最准确的计算,价格相除取对数
ARCH
EWMA
GARCH
是mean-reverting的model
计算出来的volatility与长期方差项对比
simulation process
monte carlo
几何布朗运动GBM
estimate correlation
多项资产的correlation是copula
gaussian copula
t copula
金融危机是选择这个copula,肥尾分布
foundation of risk management
risk management
risk measurement
expected return 的distribution来度量risk
absolute risk用SD来度量,relative risk用相对benchmark的tracking error来度量
evaluation of the risk measurement process
portfolio construction
sharp ratio,度量absolute risk adjusted return
information ratio,度量relative risk adjusted return
tracking error volatility(TEV)=SD1^2-2correlated coefficiency*SD1*SD2+SD2^2
treynor ratio,度量不能被diversify的systematic risk adjusted return
asset pricing theory
CAPM
APT
Value of risk management
financial disaster
kinder peabody
artificial profit
rogue trader
allied irish bank
imaginary trade
rogue trader
no trade confirmation
barings
speculate strategy
dual role
rogue trader
LTCM
russia unexpected default
insufficient equity and cash flow crisis
model risk
metallgesellschaft
german accounting
liquidity problem
market shift to contango
sumimoto
speculate strategy
lack of supervisor
Valuation and Risk Model
VaR
两种形式
dollar
percentage
不同时间VaR变换
calculation method
linear model
delta-normal,只考虑一阶矩,高阶矩不考虑
full price
historical method/monte carlo method
估计volatility
historical
parametric
EWMA
GARCH
nonparametric
hybrid
implied volatility
putting VaR to work
存在分布假设,在crisis的时候correlation显著上升
stress testing
worst case scenario
measure financial risk
mean-variance framework
对分布有假设
expected shortfall
在x quantiles外的平均损失率
对VaR的补充
scenario analysis
valuation
fix income valuation
option valuation
binomial tree
计算option到期的payoff
计算risk neutral up/down
计算probability of up/down
probability weighted net gain 连续复利贴现到当期
不断缩短时间周期就得到BSM
BSM
首先标的资产价格符合lognormal distribution
收益率服从正态分布
收藏
收藏
0 条评论
下一页