第十四课:Sampling and Monte Carlo Simulation


Monte Carlo simulation—>Inferential statistics(random sample tends to exhibit the same properties as the population)

Bernoulli’s law (law of large numbers)

Repeated independent tests with the same actual probability, p Chance that fraction of times outcome  occurs converges to p as numbers trials goes to infinity.)

第十五课:Statistical Thinking


How many samples are needed to have confidence in results?

variance — measure of how much spread there is in possible outcomes.

standard deviation标准偏差测量 — the fraction of values close to mean (计算公式如下)formula15-1


Coefficient of variation 变异系数– standard deviation by the mean (如果平均数接近0,细微的改变也会造成巨大变化,要三思)

If < 1,  low variance

变异系数can’t be used for confidence intervals, 标准差可以

normal distribution正态分布 peaks at mean, falls of symmetrically


1) Nice mathematical properties

2) Many naturally occurring examples

特性: Characterized by mean & standard deviation

confidence intervals – range likely contain unknown value & confidence level that value lies within range

Empirical rule 经验法则

1) 68% of data within 1 standard deviation of mean

2) 95% 2 sd

3)99.7% 3 sd

民意调查如何获知标准差?Standard error 少数人调查推断多数人

p = % sampled, n = sample size

SE = ((p*(100-p)/n))**0.5    //**表示幂



电子邮件地址不会被公开。 必填项已用*标注