IB数学AI HL 统计 概率 正态分布 假设检验
IB数学 Applications and Interpretation HL 的统计与概率部分，在卷一与卷二中占比很高，而且题目往往把公式套用与机考操作捆在一起考。本文把描述统计、概率、离散随机变量、二项分布、正态分布、假设检验与相关回归七个板块的核心公式、典型问法、计算器命令和易错点串成一条线，帮助你在考前建立完整的知识框架，并把方法分稳稳拿到手。
Statistics and probability take a large share of both papers in IB Maths Applications and Interpretation HL, and the questions usually combine formula recall with graphic display calculator technique. This guide walks through descriptive statistics, probability, discrete random variables, the binomial distribution, the normal distribution, hypothesis testing and correlation, with the formulas, the standard question wording, the calculator commands and the mistakes that cost the most marks.
一 描述统计与数据呈现
描述统计是整个统计模块的地基，考试要求你会算均值 mean、中位数 median、众数 mode、上下四分位数 quartiles、四分位距 IQR、极差 range 与标准差 standard deviation。最关键的一条是分清总体标准差 σ 与样本标准差 s(n-1)：题目说 population 时用除以 n 的版本，题目说 sample 或 estimate 时用除以 n-1 的版本，选错一个分母整题答案分全失。箱线图 box-and-whisker plot 的异常值判定规则是小于 Q1 减 1.5 倍 IQR，或大于 Q3 加 1.5 倍 IQR 的数据点须单独标为 outlier。线性变换的性质也必须熟：每个数据加常数 c 时均值加 c 而标准差不变，每个数据乘 k 时均值乘 k 而标准差乘 k 的绝对值。
Descriptive statistics is the foundation of the whole topic. You must be able to compute the mean, median, mode, quartiles, interquartile range, range and standard deviation, and above all to tell the population standard deviation from the sample standard deviation, since the two differ only in the divisor and the question always tells you which one to use. The outlier rule says that any value below the lower quartile minus one point five times the interquartile range, or above the upper quartile plus one point five times the interquartile range, must be drawn as a separate point on the box plot. Remember how linear transformations behave: adding a constant to every value shifts the mean by the same constant and leaves the standard deviation unchanged, while multiplying every value by k multiplies the mean and the standard deviation by k.
二 概率基础与条件概率
概率的加法公式为 P(A 或 B) 等于 P(A) 加 P(B) 减去 P(A 且 B)，当两事件互斥 mutually exclusive 时交集概率为零，公式退化为简单相加。条件概率的定义是 P(A 给定 B) 等于 P(A 且 B) 除以 P(B)，注意分母是条件事件本身的概率。独立性的判定有两条等价路径：一是 P(A 且 B) 等于 P(A) 乘 P(B)，二是 P(A 给定 B) 等于 P(A)。维恩图 Venn diagram 适合处理两个事件的重叠与补集，树状图 tree diagram 适合分阶段的试验，每一层的分支概率相加必须等于一，这是检查列式是否出错最快的办法。反向求条件概率时，用 P(B 给定 A) 等于 P(A 给定 B) 乘 P(B) 除以 P(A)，这就是贝叶斯式推理的基本形态。
The addition rule states that the probability of A or B equals the probability of A plus the probability of B minus the probability of both. For mutually exclusive events the last term is zero and the rule collapses into a simple sum. Conditional probability is defined as the probability of A given B, which equals the probability of both divided by the probability of B, so watch the denominator carefully. Two events are independent exactly when the probability of both equals the product of the individual probabilities, which is equivalent to saying that conditioning on one does not change the probability of the other. Use a Venn diagram when you need to see overlap and complement, and a tree diagram when the experiment has stages; the branches at each stage must always sum to one, which is the fastest way to check your working. To reverse a conditional probability, use the multiplication rearrangement of the definition.
三 离散随机变量与期望方差
离散随机变量 X 的概率分布表必须满足两条合法性条件：每个概率非负，且所有概率之和恰好等于一。期望的计算是 E(X) 等于各取值乘以对应概率后求和；方差用 Var(X) 等于 E(X 平方) 减去 E(X) 的平方，这个变形公式比定义式更好算，因为不必先求离差。标准差就是方差开平方。线性变换的规律与数据一节完全一致：整体加常数时期望整体平移而方差不变，整体乘 k 时期望乘 k 而方差乘 k 平方。判断一个游戏是否公平 fair game，标准就是期望收益等于零；若题目给出参与费用，要先把费用计入净收益再比较。机考中先用列表输入取值与概率，再用统计功能一键求出均值与标准差，比手算快很多。
A discrete random variable needs a valid probability distribution: every probability lies between zero and one, and the total is exactly one. The expectation is found by multiplying each value by its probability and adding the results, while the variance is most easily obtained as the mean of the squares minus the square of the mean. The standard deviation is simply the square root of the variance. Linear transformations behave exactly as they do for raw data: adding a constant shifts the expectation and leaves the variance untouched, whereas multiplying by k multiplies the expectation by k and the variance by k squared. A game is fair when the expected net gain is zero, so remember to subtract any entry fee before you compare. On the calculator, enter the values and probabilities as lists and use the one variable statistics routine rather than doing the arithmetic by hand.
四 二项分布的条件与计算
二项分布适用于满足四个条件的场景：试验次数 n 固定、每次试验只有成功与失败两种结果、成功概率 p 在每次试验中保持不变、各次试验相互独立。概率公式为 P(X 等于 r) 等于组合数 C(n, r) 乘 p 的 r 次方 乘 (1-p) 的 (n-r) 次方。均值 E(X) 等于 np，方差 Var(X) 等于 np(1-p)，这两个公式在选择题中可以直接使用。机考命令要写清楚：binompdf 求单点概率，binomcdf 求累积概率；遇到"至少"或"多于"一类问法，写成一减去反面累积概率，比逐项相加快且不易漏项。题目若说 without replacement 即不放回抽样，就不再是二项分布，这是最常见的陷阱之一。
The binomial distribution applies when four conditions hold: the number of trials is fixed, each trial has only two possible outcomes, the probability of success stays constant from trial to trial, and the trials are independent. The probability of exactly r successes uses the binomial coefficient multiplied by the success probability raised to r and the failure probability raised to n minus r. The mean equals n times p and the variance equals n times p times one minus p, and both can be quoted directly in multiple choice questions. Write the calculator command and its arguments explicitly, using the probability function for a single value and the cumulative function for a range. Questions asking for at least some value are almost always fastest as one minus the complementary cumulative probability. If the question says without replacement, the trials are no longer independent and the binomial model does not apply.
五 正态分布与标准化
正态分布 X 服从 N(μ, σ 平方) 时，完全由均值与方差两个参数决定，密度曲线关于均值对称，曲线下总面积为 1。标准化公式 z 等于 (x 减 μ) 除以 σ，把任意正态变量转化为标准正态变量，再查表或用计算器求概率。反向问题给出的是概率，要求的是分位点，用反正态函数 inversenorm 直接求解，注意区分单尾与双尾，双尾时要考虑两侧的对称性。二项分布的正态近似条件是 np 与 n(1-p) 同时大于 5，近似时必须做连续性修正 continuity correction，把离散的整数端点向外扩 0.5。计算器求正态概率时，下界用 -1E99 代替负无穷，上界用 1E99 代替正无穷，否则命令会报错。
A normal distribution is completely determined by its mean and its variance. Its density curve is symmetric about the mean and encloses a total area of one, which is why probabilities are always read as areas. Standardising with z equal to x minus the mean divided by the standard deviation converts any normal variable into the standard normal variable, from which probabilities follow. The reverse question gives you a probability and asks for the value, so use the inverse normal function, and be careful to distinguish a one tailed from a two tailed requirement because symmetry matters in the latter case. A binomial variable may be approximated by a normal distribution when both n times p and n times one minus p exceed five, but then a continuity correction of half a unit must be applied at the integer boundaries. When you enter normal probabilities on the calculator, use an extreme value such as minus one times ten to the ninety ninth for the open lower bound.
六 假设检验与两类错误
假设检验的答题结构是四件套：先写零假设 H0 与备择假设 H1，再声明显著性水平，通常取百分之五或百分之一，然后算出 p 值，最后与水平比较得出结论。只有当备择假设指明了方向时才用单尾检验 one-tailed test，双侧检验要把显著性水平对半分给两端。p 值小于显著性水平时拒绝 H0，但结论句必须回到题目的语境里写，例如直接说明有足够证据支持平均分发生了变化，而不是只写拒绝零假设。两类错误要分清：第一类错误 Type I error 是 H0 本来为真却被拒绝，第二类错误是 H0 本来为假却未被拒绝。AI HL 还会考 t 检验与卡方拟合优度检验，前者自由度是 n 减一，后者自由度是类别数减一，前提是每个期望频数不小于五。
A hypothesis test answer has four parts: the null hypothesis, the alternative hypothesis, the stated significance level, and the p value, followed by a contextual conclusion. Use a one tailed test only when the alternative hypothesis states a direction, and split the significance level between the two tails for a two tailed test. Reject the null hypothesis when the p value is smaller than the significance level, but always write the conclusion in the words of the question rather than repeating the abstract decision. A Type I error means rejecting a null hypothesis that is actually true, while a Type II error means failing to reject one that is false. The t test and the chi squared goodness of fit test also appear in this course, with degrees of freedom of n minus one and the number of categories minus one respectively, and the chi squared test requires every expected frequency to be at least five.
七 相关与回归
皮尔逊相关系数 r 衡量线性关系的强弱与方向，取值在负一与正一之间，r 的绝对值越接近一线性关系越强。斯皮尔曼等级相关系数 rs 用于单调但非线性的数据，或用于序数数据，计算时先把数据转换成排名再套用公式。回归直线用最小二乘法得到，斜率的正负与相关系数的正负必然一致，这是快速自查的方法。使用回归直线时只能在数据范围内插值 interpolation，超出数据范围的外推 extrapolation 不可靠，答题时要明确说明这一点。另外 r 接近零只说明没有线性关系，不代表两个变量之间毫无关系，可能存在明显的非线性关联，这句话在概念题中经常是得分的要点。
The Pearson correlation coefficient measures the strength and direction of a linear relationship and takes values between minus one and one, with values near either extreme indicating a strong linear association. The Spearman rank correlation coefficient is used for monotonic but non linear relationships and for ordinal data, and it is computed from the ranks of the observations rather than their raw values. A least squares regression line always has a slope with the same sign as the correlation coefficient, which is a quick way to check your answer. A regression line should only be used for interpolation inside the range of the observed data; extrapolation beyond that range is unreliable and should be stated as such. Finally, a correlation coefficient near zero means there is no linear relationship, not that there is no relationship at all, and that distinction is a frequent source of concept marks.
八 考试技巧与常见失分点
第一，题目要求三位有效数字时，中间步骤要保留更多位，只在最后一步做舍入，过早舍入会累积误差。第二，机考过程分与答案分分开给，务必把计算器命令与参数写出来，只写一个数字可能拿不到方法分。第三，假设检验按四件套书写，缺任何一件都会扣分。第四，结论句必须带上单位与题目语境。第五，树状图与箱线图先在草稿纸上画出来，再套公式，比凭记忆直接列式更稳。第六，考前把整套公式表按板块默写一遍，尤其是方差的两个变形、正态标准化、二项分布的两个参数以及自由度规则。
First, when the question demands three significant figures, carry more digits through the working and round only at the very end, because premature rounding accumulates error. Second, method marks are awarded separately from answer marks, so always write down the calculator command and its arguments instead of a bare number. Third, structure a hypothesis test with all four required parts, since any missing element loses marks. Fourth, keep units and the context of the question in your concluding sentence. Fifth, sketch tree diagrams and box plots on paper before substituting into formulas, which is more reliable than assembling the algebra from memory. Sixth, before the exam rewrite the formula booklet section by section, paying particular attention to the two forms of the variance, standardisation for the normal distribution, the two parameters of the binomial model and the degrees of freedom rules.
九 常用公式速查
考前把下面这一组关系默写一遍：方差等于平方的期望减去期望的平方；标准差等于方差开方；正态标准化 z 等于 (x 减 μ) 除以 σ；二项分布的期望是 np、方差是 np(1-p)；二项分布正态近似的条件是 np 与 n(1-p) 均大于 5，并且必须做连续性修正；卡方拟合优度检验的期望频数不得小于 5。把这些关系与数据线性变换的两条规律放在一起记，你会发现它们其实是同一套方差代数在不同场景下的重复出现，理解了这一点比死背公式更抗遗忘。
Before the exam, write out this short list from memory: the variance equals the mean of the squares minus the square of the mean; the standard deviation is the square root of the variance; standardisation takes x minus the mean over the standard deviation; the binomial mean is n times p and its variance is n times p times one minus p; the normal approximation to the binomial requires both n times p and n times one minus p to exceed five and a continuity correction; and the chi squared test requires expected frequencies of at least five. Notice that these rules and the two laws of linear transformation are really the same variance algebra appearing in different settings, and understanding that repetition is far more durable than memorising each formula separately.
海外直订Higher Mathematics for AQA GCSE 高等数学AQA GCSE¥645.0📱 去京东购买京东精选
需要一对一辅导？Need One-on-One Tutoring?
A-Level / GCSE / IB / KS2-KS3 全科辅导
16621398022 同微信
Visit aleveler.com
或截图后用「扫一扫 → 相册」选择本图