113 年 國立臺灣大學商學研究所甲組《統計學(H)》

📄 試題原卷 免費註冊後即可對照原始考卷 PDF免費註冊

第 1 題

Please choose the most suitable answer among all alternatives for each question and clearly mark it on the answer sheet. Each
question is worth 5 points.

  1. Which descriptive statistic is least affected by extremely small or extremely large data values?
    A) Mean
    B) Median
    C) Range
    D) Mean and Median
    E) Median and Range

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查敘述統計中各項指標對極端值的敏感度。

  • 平均數 (Mean):受所有資料點影響,極端值會顯著拉高或拉低平均數。
  • 中位數 (Median):是將資料排序後位於中間的數值,不受極端值影響,因為它只關乎排序的位置。
🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 2 題

  1. What is the relationship among the mean, median, and mode in a right-skewed (or positively skewed) distribution?
    A) The mean is typically the smallest value.
    B) The median is typically the smallest value.
    C) The mode is typically the smallest value.
    D) They are all equal.
    E) Cannot conclude without further information.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查偏態分配(skewed distribution)中,平均數、中位數與眾數的相對位置關係。

右偏(或稱正偏,right-skewed / positively skewed)的分配,其尾部向右延伸,表示有少數極大的數值將平均數往右(較大值)拉。此時,眾數(mode)代表出現頻率最高的點,通常在峰值處;中位數(median)則是在排序後的中間值;

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 3 題

  1. Suppose there are two distributions, A and B. A is a normal distribution with a mean of 4 and a standard deviation of 3. B is a normal distribution with a mean of 5 and a standard deviation of 2. Which of the following statements is false?
    A) The width of distribution A is wider than that of distribution B.
    B) The mean, median, and mode are equal to 4 for distribution A and 5 for distribution B.
    C) The frequency of values peaks at the mean for both distributions.
    D) 100% of the values fall between ±3 standard deviations for both distributions A and B.
    E) All of the statements above are correct.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查常態分配的性質,以及標準差與分配寬度的關係。

給定:
分配 A:常態分配,μA=4\mu_A = 4, σA=3\sigma_A = 3
分配 B:常態分配,μB=5\mu_B = 5, σB=2\sigma_B = 2

我們逐一檢視各個敘述:

A) The width of distribution A is wider than that of distribution B.
分配的寬度通常由標準差 (σ\sigma) 來衡量。標準差越大,分配越寬(越扁平)。
σA=3\sigma_A = 3, σB=2\sigma_B = 2。因為 σA>σB\sigma_A > \sigma_B,所以分配 A 的寬度比分配 B 寬。此敘述為真。

B) The mean, median, and mode are equal to 4 for distribution A and 5 for distribution B.
對於常態分配,平均數 (μ\mu)、中位數 (median) 和眾數 (mode) 是相等的。
對於分配 A,μA=4\mu_A = 4,所以其平均數、中位數和眾數均為 4。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 4 題

  1. The wait time at a specific stoplight follows a uniform distribution ranging from zero to five minutes. What is the probability of having to wait more than 120 seconds at the stoplight?
    A) 0.2
    B) 0.4
    C) 0.6
    D) 0.8
    E) 1.0

登入後即可作答並保存紀錄。

這一題的完整詳解

核心觀念

本題考查連續型均勻分配(uniform distribution)。

若隨機變數 XX 服從區間 [a,b][a,b] 上的均勻分配,則其機率密度在區間內相同,事件發生的機率可用「區間長度比」計算:

P(c<X<d)=d−cb−aP(c<X<d)=\frac{d-c}{b-a}

本題中等待時間介於 00 分鐘至 55 分鐘,因此:

X∼U(0,5)X\sim U(0,5)

題目要求等待超過 120120 秒的機率,必須先將秒數換算成分鐘:

120 秒=2 分鐘120\text{ 秒}=2\text{ 分鐘}

解題方法

等待超過 22 分鐘,代表 X>2X>2。在總等待區間 [0,5][0,5] 中,符合條件的區間是 $(2,5],其長度為:

5−2=35-2=3

總區間長度為:

5−0=55-0=5

因此:

P(X>2)=5−25−0=35=0.6P(X>2)=\frac{5-2}{5-0} =\frac{3}{5} =0.6

由於均勻分配是連續型分配,單一點的機率為 00,所以 P(X>2)P(X>2) 與 P(X≥2)P(X\geq 2) 相同。

選項分析

  • A) 0.2:錯誤。
    0.20.2 是 1/51/5,相當於只取總區間的五分之一,並非等待超過 $2
🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 5 題

  1. Why is the central limit theorem important in statistics?
    A) Because for a large sample size n, it asserts the population is approximately normal.
    B) Because for any population, it asserts the sampling distribution of the sample mean is approximately normal, regardless of
    the shape of the population.
    C) Because for a large sample size n, it asserts the sampling distribution of the sample mean is approximately normal,
    regardless of the shape of the population.
    D) Because for any sample size n, it asserts the sampling distribution of the sample mean is approximately normal.
    E) None of the above.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查中央極限定理 (Central Limit Theorem, CLT) 的重要性與內容。

中央極限定理的核心思想是:
當從任何一個母體(無論其原始分布為何種形狀)中抽取足夠大的樣本(通常認為 n≥30n \ge 30 是一個足夠大的樣本量),這些樣本的平均數的抽樣分配(sampling distribution of the sample mean)將近似於常態分配。

我們來逐一分析選項:
A) Because for a large sample size n, it asserts the population is approximately normal.
CLT 說明的是樣本平均數的抽樣分配近似常態,而不是母體本身近似常態。這個敘述是錯誤的。

B) Because for any population, it asserts the sampling distribution of the sample mean is approximately normal, regardless of the shape of the population.
這個敘述前半句「for any population」是正確的,CLT 適用於任何母體。後半句「it asserts the sampling distribution of the sample mean is approximately normal, regardless of the shape of the population」也是正確的。但是,CLT 的一個重要條件是樣本必須足夠大。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 6 題

  1. A company is recruiting management trainees for entry-level marketing positions. Based on historical data, approximately 20% are expected to remain employed after six months. If the company has recently hired five trainees, what is the approximate probability that exactly three of them will still be employed at the end of six months?
    A) 0.005
    B) 0.008
    C) 0.05
    D) 0.2
    E) 0.5

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查二項分配 (binomial distribution) 的機率計算。

題目給定:

  1. 每次試驗(每位新進員工)有兩種結果:留任或離職。
  2. 留任的機率(成功機率)p=0.20p = 0.20。
  3. 離職的機率(失敗機率)q=1−p=1−0.20=0.80q = 1 - p = 1 - 0.20 = 0.80。
  4. 總共進行了 n=5n = 5 次獨立試驗(共聘用 5 位員工)。
  5. 我們想求的是恰好有 k=3k = 3 位員工留任的機率。

這符合二項分配的條件,其機率質量函數 (PMF) 為:
P(X=k)=(nk)pkqn−kP(X=k) = \binom{n}{k} p^k q^{n-k}

將題目數值代入:
n=5n = 5, k=3k = 3, p=0.2p = 0.2, q=0.8q = 0.8

P(X=3)=(53)(0.2)3(0.8)5−3P(X=3) = \binom{5}{3} (0.2)^3 (0.8)^{5-3}
P(X=3)=(53)(0.2)3(0.8)2P(X=3) = \binom{5}{3} (0.2)^3 (0.8)^2

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 7 題

  1. For a given population proportion, which of the following statements is true regarding the width of a confidence interval for the population proportion?
    A) It is narrower for a lower confidence level than for a higher confidence level.
    B) It is wider for a larger sample size than for a smaller sample size.
    C) It is narrower for a larger sample proportion than for a smaller sample proportion.
    D) It is wider for a larger sample proportion than for a smaller sample proportion.
    E) None of the above is correct.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查信賴區間 (confidence interval) 的寬度與其相關因素的關係,特別是針對母體比例 (population proportion) 的信賴區間。

母體比例 p^\hat{p} 的信賴區間公式通常為:
p^±zα/2p^(1−p^)n\hat{p} \pm z_{\alpha/2} \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}
其中,信賴區間的寬度 (width) 為 2×margin of error=2×zα/2p^(1−p^)n2 \times \text{margin of error} = 2 \times z_{\alpha/2} \sqrt{\frac{\hat{p}(1-\hat{p})}{n}}。

我們逐一檢視各個選項:

A) It is narrower for a lower confidence level than for a higher confidence level.
信賴水準越高 (例如 99% vs 95%),α\alpha 越小,α/2\alpha/2 越小,因此 zα/2z_{\alpha/2} 越大。zα/2z_{\alpha/2} 是影響寬度的因子。所以,較高的信賴水準對應較大的 zα/2z_{\alpha/2},導致信賴區間更寬。反之,較低的信賴水準對應較小的 zα/2z_{\alpha/2},導致信賴區間更窄。此敘述為真。

B) It is wider for a larger sample size than for a smaller sample size.
信賴區間的寬度與樣本大小 nn 成反比(在分母的平方根下)。若樣本大小 nn 越大,則 p^(1−p^)n\sqrt{\frac{\hat{p}(1-\hat{p})}{n}} 的值越小,信賴區間的寬度越窄。此敘述為偽。

C) It is narrower for a larger sample proportion than for a smaller sample proportion.
信賴區間的寬度與 p^(1−p^)\sqrt{\hat{p}(1-\hat{p})} 成正比。函數 f(p)=p(1−p)f(p) = p(1-p) 在 p=0.5p=0.5 時達到最大值 (0.25),在 p=0p=0 或 p=1p=1 時達到最小值 (0)。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 8 題

  1. The head of a research team claims that he can accurately determine whether a person has a medical background or an engineering background based on their problem-solving approach. When presented with one individual and asked to identify their background (either medical or engineering), the team lead treats this as a hypothesis test with the null hypothesis being that the person has a medical background and the alternative that the person has an engineering background. Which of the following statements illustrates a Type II error?
    A) Identifying the person as having an engineering background when, in fact, the person has a medical background.
    B) Identifying the person as having a medical background when, in fact, the person has a medical background.
    C) Identifying the person as having a medical background when, in fact, the person has an engineering background.
    D) Identifying the person as having an engineering background when, in fact, the person has an engineering background.
    E) None of the above.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查統計檢定中的第二類錯誤 (Type II error)。

假設檢定的設定如下:
虛無假設 H0H_0: 此人有醫學背景 (Medical background)。
對立假設 H1H_1: 此人有工程背景 (Engineering background)。

我們知道兩種錯誤的定義:

  • 第一類錯誤 (Type I error):當虛無假設 H0H_0 為真時,拒絕 H0H_0。
  • 第二類錯誤 (Type II error):當虛無假設 H0H_0 為偽(即對立假設 H1H_1 為真)時,未能拒絕 H0H_0。

現在我們來分析每個選項:

A) Identifying the person as having an engineering background when, in fact, the person has a medical background.
這表示我們認為 H1H_1 為真(工程背景),但實際上 H0H_0 為真(醫學背景)。這是第一類錯誤 (Type I error)。

B) **Identifying the person as having a medical background when, in fact, the person has a m

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 9 題

  1. Which of the following does not influence the standard error of the regression slope?
    A) The spread around the regression line
    B) The spread of x values
    C) The sample size
    D) The critical value
    E) All of these affect the standard error

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查簡單線性迴歸中,迴歸斜率估計值的標準誤 (standard error of the regression slope) 的影響因素。

簡單線性迴歸模型為 yi=β0+β1xi+ϵiy_i = \beta_0 + \beta_1 x_i + \epsilon_i。
迴歸斜率 β1\beta_1 的估計值為 β^1\hat{\beta}_1。
β^1\hat{\beta}_1 的標準誤 (SE(β^1\hat{\beta}_1)) 的公式為:
SE(β^1)=se2∑i=1n(xi−xˉ)2SE(\hat{\beta}_1) = \sqrt{\frac{s_e^2}{\sum_{i=1}^n (x_i - \bar{x})^2}}
其中,se2s_e^2 是殘差變異數 (residual variance),它估計了誤差項 ϵi\epsilon_i 的變異數 σ2\sigma^2。
se2=SSEn−2=∑i=1n(yi−y^i)2n−2=∑i=1nei2n−2s_e^2 = \frac{SSE}{n-2} = \frac{\sum_{i=1}^n (y_i - \hat{y}_i)^2}{n-2} = \frac{\sum_{i=1}^n e_i^2}{n-2}。

我們來分析公式中的各個部分與選項的關聯:

A) The spread around the regression line (殘差平方和 SSE 或殘差變異數 se2s_e^2)
se2s_e^2 在分子上,所以 se2s_e^2 越大(即殘差越大,點越分散在迴歸線周圍),SE(β^1)SE(\hat{\beta}_1) 就越大。此選項影響標準誤。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 10 題

  1. To evaluate the effectiveness of a new exercise program, a random sample of 25 individuals is chosen from a population of adults engaged in the program. Each participant's weight is measured both before and after the program. Assuming that the population of differences in weight before and after the program follows a normal distribution, the mean decrease in weights for the 25 participants is determined to be 2.5 pounds, with a standard deviation of differences of 3.2 pounds. What conclusion can be drawn about the effectiveness of the exercise program based on the test at the 0.05 level of significance?
    A) The exercise program is not effective.
    B) The exercise program is effective.
    C) The sample size is relatively small to make a conclusion.
    D) A conclusion cannot be drawn without calculating the p-value.
    E) None of the above.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查配對樣本 tt 檢定 (paired samples t-test) 的應用。

題目給定:

  • 樣本大小 n=25n = 25。
  • 樣本平均差異(體重減輕量)dˉ=2.5\bar{d} = 2.5 磅。
  • 樣本差異的標準差 sd=3.2s_d = 3.2 磅。
  • 假設母體差異服從常態分配。
  • 顯著水準 α=0.05\alpha = 0.05。

我們要檢定的是運動計畫的有效性。有效性意味著體重有顯著減輕。
因此,虛無假設 H0H_0 和對立假設 H1H_1 設定如下:
H0H_0: 運動計畫沒有效果,即平均體重減輕量為 0 或更少。μd≤0\mu_d \le 0 (其中 μd\mu_d 是母體平均差異,定義為體重減輕量)。
H1H_1: 運動計畫有效,即平均體重減輕量大於 0。μd>0\mu_d > 0。

這是一個單尾檢定 (one-tailed test)。

檢定統計量 tt 的計算公式為:
t=dˉ−μd0sd/nt = \frac{\bar{d} - \mu_{d0}}{s_d / \sqrt{n}}
其中,μd0\mu_{d0} 是虛無假設下的母體平均差異,在此為 0。

將數值代入:
t=2.5−03.2/25=2.53.2/5=2.50.64t = \frac{2.5 - 0}{3.2 / \sqrt{25}} = \frac{2.5}{3.2 / 5} = \frac{2.5}{0.64}

計算 tt 值:
t=2.50.64=25064=12532≈3.906t = \frac{2.5}{0.64} = \frac{250}{64} = \frac{125}{32} \approx 3.906

接著,我們需要確定自由度 (degrees of freedom, df)。對於配對樣本 tt 檢定,自由度為 df=n−1df = n - 1。
df=25−1=24df = 25 - 1 = 24。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 11 題

  1. Which of the following is a violation of one of the major assumptions of the simple regression model?
    A) The error terms are independent of each other.
    B) A histogram of the residuals forms a bell-shaped, symmetrical curve.
    C) A plot of the residual versus x forms a horizontal band pattern.
    D) As the value of x increases, the value of the error term also increases.
    E) The error terms show no pattern.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查簡單線性迴歸模型 (simple linear regression model) 的主要假設及其違反情況。

簡單線性迴歸模型的四大主要假設通常是:

  1. 線性關係 (Linearity):自變數 xx 與應變數 yy 之間的關係是線性的。
  2. 獨立性 (Independence):誤差項 ϵi\epsilon_i 之間是相互獨立的。
  3. 同質變異數 (Homoscedasticity):誤差項 ϵi\epsilon_i 的變異數是恆定的(即 Var(ϵi)=σ2\text{Var}(\epsilon_i) = \sigma^2 對所有 ii 成立)。這意味著殘差圖應該呈現一個大致的水平帶狀。
  4. 常態性 (Normality):誤差項 ϵi\epsilon_i 服從常態分配,平均值為 0。

我們來逐一檢視各選項是否違反這些假設:

A) The error terms are independent of each other.
這是迴歸模型的假設之一,而不是違反。如果誤差項不獨立(例如存在時間序列上的自相關),則會違反此假設。

B) A histogram of the residuals forms a bell-shaped, symmetrical curve.
這是對誤差項(或其估計值殘差)服從常態分配的驗證,而不是違反。如果直方圖不像鐘形曲線,則可能違反常態假設。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 12 題

  1. Which, if any, of the following statements about the chi-square test of independence is false?
    A) If rir_i is the row total for row ii and cjc_j is the column total for column jj, then the estimated expected cell frequency corresponding to row ii and column jj equals (ri)(cj)/n(r_i)(c_j)/n.
    B) The test is valid if all of the estimated cell frequencies are at least five.
    C) The chi-square statistic is based on (r−1)(c−1)(r-1)(c-1) degrees of freedom, where rr and cc denote, respectively, the number of rows and columns in the contingency table.
    D) The alternative hypothesis states that the two classifications are statistically independent.
    E) All of the other statements about the chi-square test of independence are true.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查卡方檢定獨立性 (chi-square test of independence) 的相關敘述,找出錯誤的敘述。

卡方檢定獨立性是用來檢定兩個類別變數 (categorical variables) 的分類之間是否獨立。

我們逐一檢視各個選項:

A) If rir_i is the row total for row ii and cjc_j is the column total for column jj, then the estimated expected cell frequency corresponding to row ii and column jj equals (ri)(cj)/n(r_i)(c_j)/n.
這是卡方檢定獨立性中計算期望次數 (expected frequency) 的標準公式。其中 nn 是總樣本數。此敘述為真。

B) The test is valid if all of the estimated cell frequencies are at least five.
這是卡方檢定獨立性的一個常用規則,用於確保檢定的有效性。一般要求所有單元的期望次數 (expected cell frequencies) 都大於或等於 5。若有少數單元期望次數小於 5,可以考慮合併類別或使用 Fisher's exact test。但若期望次數普遍偏低,則檢定結果可能不可靠。這個敘述的表述方式是「if all... are at least five」,這通常被視為一個充分條件,但不是必要條件。然而,在多數考題中,這個說法被視為正確的有效性條件。
更精確的說法是,期望次數普遍不應低於 5,且不應有超過 20% 的單元期望次數低於 5。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 13 題

  1. A marketer aims to test the effectiveness of the ABC filter designed to reduce harmful chemicals in drinking water. The marketer measured harmful chemical levels in 30 households' water before installing the product. Next, the same 30 households installed the ABC filter. After two weeks of continuous use, the harmful chemical levels were measured again. The comparison of harmful chemical levels before versus after using the product is an example of testing the difference between
    A) two means from independent populations.
    B) two population variances from independent populations.
    C) two population proportions.
    D) two population medians.
    E) matched pairs from two dependent populations.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查統計檢定方法的選擇,特別是針對前後測量資料的比較。

題目描述的情況:

  1. 有 30 個家庭 (households)。
  2. 在安裝 ABC 濾水器之前,測量了有害化學物質的水平。
  3. 相同的 30 個家庭安裝了濾水器。
  4. 在使用了濾水器之後,再次測量了有害化學物質的水平。
  5. 比較的是「使用產品前」和「使用產品後」的有害化學物質水平。

這種設計涉及對同一組對象(30 個家庭)進行兩次測量(安裝前和安裝後)。這表示這兩組測量值是相關聯的,是來自同一批對象在不同時間點的數據。這種情況屬於配對樣本 (paired samples) 或相依樣本 (dependent samples)。

我們需要檢定的是這兩次測量之間的差異。通常,我們檢定的是平均差異。

選項分析:
A) two means from independent populations. (來自獨立母體的兩個平均數。)
如果我們有兩組完全不同的家庭,例如 A 組家庭使用濾水器,B 組家庭不使用濾水器,然後比較 A 組和 B 組的平均化學物質水平,那這才是獨立樣本。但在本題中,是同一組家庭前後測量,所以樣本是相依的。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 14 題

For Questions 14-16, please read the scenario and output below.
To compare the performance characteristics of four brands of space heaters, a researcher bought all the heaters they could find
from these brands and tested them. For performance, the researcher measured the amount of time it took to warm up a
specific-sized room by 5 degrees from a set temperature. They also coded the characteristics of the space heaters. The
researcher first conducted a one-way ANOVA on the space heaters' performance among the four brands. The output is given
below.

ANOVA table
Source SS df MS F p-value
Treatment 6.000 3 1.9998 18.85 3.46E-05
Error 1.713 16 0.1061
Total 7.713 19

Post hoc analysis
Tukey simultaneous comparison t-values (d.f. = 16)
Brand 2 Brand 4 Brand 1 Brand 3
Brand 2 1.40 2.28 2.58 2.95
Brand 3 2.28 4.27 5.38 1.35
Brand 4 2.58 5.38 3.07 1.63
Brand 1 2.95 1.35 1.63 7.09

Critical values for experimentwise error rate:
0.05 2.91
0.01 3.76

  1. Based on the output, at a significance level of .05, we would conclude that
    A) brand 1 differs from brand 3, and brand 3 differs from brand 2, while the rest of the space heater pairs do not differ
    from each other in terms of their performance.
    B) brand 1 differs from brand 3, and brand 2 differs from brands 1, 3, and 4, while the rest of the space heater pairs do
    not differ from each other in terms of their performance.
    C) only brand 2 differs from the other three brands, while the rest of the space heater pairs do not differ from each other
    in terms of their performance.
    D) all four brands of space heaters differ from each other in terms of their performance.
    E) none of the four brands of space heaters differ from each other in terms of their performance.

登入後即可作答並保存紀錄。

這一題的完整詳解

核心觀念

本題考查單因子變異數分析(one-way ANOVA)後的 Tukey 多重比較。

ANOVA 的虛無假設為:

H0:μ1=μ2=μ3=μ4H_0:\mu_1=\mu_2=\mu_3=\mu_4

若整體 ANOVA 顯著,只能表示四個品牌的平均暖房時間不全相同,不能直接判斷究竟是哪幾組不同。因此必須利用事後比較(post hoc analysis)逐一檢定各品牌平均數差異。

Tukey simultaneous comparison 使用實驗族錯誤率(experimentwise error rate)控制多重比較下的整體第一類錯誤率。本題在顯著水準 α=0.05\alpha=0.05 下,臨界值為:

tcritical=2.91t_{\text{critical}}=2.91

判斷原則為:

∣tTukey∣>2.91⟹兩品牌平均性能有顯著差異|t_{\text{Tukey}}|>2.91 \quad\Longrightarrow\quad \text{兩品牌平均性能有顯著差異}

否則視為沒有顯著差異。


解題方法

首先觀察 ANOVA 表:

F=18.85,p-value=3.46×10−5F=18.85,\qquad p\text{-value}=3.46\times 10^{-5}

因為:

3.46×10−5<0.053.46\times 10^{-5}<0.05

所以拒絕整體平均數相同的虛無假設,表示四個品牌中至少有一組平均暖房時間不同。

接著進行 Tukey 事後比較。將各品牌配對比較的 tt 值與 0.050.05 的臨界值 2.912.91 比較,可得到:

  • Brand 1 與 Brand 3:t=7.09>2.91t=7.09>2.91,顯著不同。
  • Brand 2 與 Brand 1:t=4.27>2.91t=4.27>2.91,顯著不同。
  • Brand 2 與 Brand 3:t=5.38>2.91t=5.38>2.91,顯著不同。
  • Brand 2 與 Brand 4:t=3.07>2.91t=3.07>2.91,顯著不同。
  • Brand 1 與 Brand 4:t=1.35<2.91t=1.35<2.91,無顯著差異。
  • Brand 3 與 Brand 4:t=1.63<2.91t=1.63<2.91,無顯著差異。
🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 15 題

  1. How many heaters did the researcher test in this research?
    A) 17
    B) 19
    C) 20
    D) 21
    E) 38

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查從 ANOVA 表格中判斷樣本數。

ANOVA 表格中的「Source」欄位顯示了變異的來源,包括「Treatment」(處理,即品牌)和「Error」(誤差)。
「df」欄位代表自由度 (degrees of freedom)。

  • 處理自由度 (df Treatment):代表組別數減一。題目中提到有四個品牌的空間加熱器,所以組別數是 4。因此,dfTreatment=組別數−1=4−1=3df_{Treatment} = \text{組別數} - 1 = 4 - 1 = 3。這與表格中的值 3 相符。
  • 誤差自由度 (df Error):代表總樣本數減去組別數。
🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 16 題

  1. To further understand the space heaters' performance, in addition to the brand, the researcher would like to add voltage as
    a predictor. Hence, they decided to conduct a multiple regression model. How many independent variables should they
    create for the brand in this multiple regression model?
    A) One
    B) Two
    C) Three
    D) Four
    E) Five

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查在建立複迴歸模型 (multiple regression model) 時,如何處理類別變數 (categorical variable) 作為自變數。

題目提到,除了品牌 (brand) 之外,研究者還想加入電壓 (voltage) 作為預測因子。
品牌是一個類別變數,有四個類別(四個品牌)。
在複迴歸模型中,類別變數需要轉換成數值變數才能使用。常見的方法是使用虛擬變數 (dummy variables) 或稱指示變數 (indicator variables)。

對於一個有 kk 個類別的類別變數,我們通常需要建立 k−1k-1 個虛擬變數。
在本題中,品牌有 4 個類別(Brand 1, Brand 2, Brand 3, Brand 4)。

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 17 題

For Questions 17-20, please read the scenario below.
A consultant hired by a chain restaurant focusing on eco-friendly practices has developed a multiple regression model for the
chain restaurants' performances for their branches. In this multiple regression model, the dependent variable, performances, is
the branch sales in thousands of dollars. For example, a data entry of 25 for the dependent variable indicates a sales of $25,000.
The dataset the consultant was given also included branch and market factors, such as the adaptation of eco-friendly practices,
branch size, employee tenure, and population density of surrounding areas.

  1. The consultant first tested a multiple regression model with the eco-friendly variable as one of the indicators (X1 = 0 if
    not adopted and X1 = 1 if adopted). The output of this multiple regression model shows that the coefficient for this
    variable (X1) is -1.6. The t test showed that X1 was significant at a = .05. This result implies that for companies adopting
    and not adopting sustainable practices,
    A) on average, the branches that adopted eco-friendly practices made 1,600lessthanthosethatdidnotadopteco−friendlypractices.B)onaverage,thebranchesthatdidnotadopteco−friendlypracticesmade1,600 less than those that did not adopt eco-friendly practices. B) on average, the branches that did not adopt eco-friendly practices made 1,600 less than those that adopted
    eco-friendly practices.
    C) on average, the branches that adopted eco-friendly practices made 1.6 times less than those that did not adopt
    eco-friendly practices.
    D) on average, the branches that did not adopt eco-friendly practices made 1.6 times less than those that adopted
    eco-friendly practices.
    E) on average, the performances were not affected by whether the branches adopted eco-friendly practices or not.

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查複迴歸模型中,虛擬變數 (dummy variable) 係數的解釋。

模型中:

  • 應變數 (Dependent Variable):Performance (branch sales in thousands of dollars)。
  • 自變數 (Independent Variable):Eco-friendly practice adoption (X1)。
    • X1=0X_1 = 0:未採用環保措施 (not adopted)。
    • X1=1X_1 = 1:已採用環保措施 (adopted)。
  • X1X_1 的迴歸係數 (β1\beta_1) = -1.6。
  • X1X_1 在 α=0.05\alpha = 0.05 下顯著。

在複迴歸模型中,當一個自變數是虛擬變數時,其係數的解釋是:與參考類別相比,該自變數為 1 的群體,在控制其他變數後,應變數的平均值會增加或減少多少。

在此情況下,參考類別是 X1=0X_1 = 0(未採用環保措施)。
X1=1X_1 = 1(已採用環保措施)的係數是 -1.6。
應變數的單位是「thousands of dollars」。

因此,係數 -1.6 表示:
已採用環保措施的店家(X1=1X_1=1),其平均業績(銷售額)比未採用環保措施的店家(X1=0X_1=0),在控制其他變數後,平均減少了 1.6 千美元。

換算成實際金額:
1.6 千美元 = 1.6×1000=16001.6 \times 1000 = 1600 美元。

選項分析:
A) **on average, the branches that adopted eco-friendly practices made

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 18 題

  1. The result on eco-friendly practice motivates the consultant to dig deeper into the dataset. While exploring, the consultant
    accidentally adds a very insignificant independent variable (an independent variable that has a very weak relationship with
    the dependent variable) to a multiple regression equation. As a result of this change, the value of the explained variation
    (SSR) will ______, the value of the multiple coefficient of determination (R²) will ______, and the calculated value
    of the F statistic will most likely ______.
    A) decrease, increase, decrease.
    B) increase, decrease, decrease
    C) increase, increase, increase
    D) increase, increase, decrease
    E) decrease, decrease, decrease

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查在複迴歸模型中,加入一個不顯著(或非常弱的)自變數對模型統計量的影響。

模型中的幾個關鍵統計量:

  • SSR (Sum of Squares due to Regression):迴歸平方和,代表由迴歸模型解釋的總變異。
  • R² (Coefficient of Determination):判定係數,代表應變數總變異中被迴歸模型解釋的比例。R2=SSRSSTR^2 = \frac{SSR}{SST},其中 SST 是總平方和。
  • F statistic:用於檢定整個迴歸模型的顯著性(即所有自變數的係數是否同時為零)。

當我們加入一個新的自變數到迴歸模型中時:

  1. SSR (Explained Variation):
    新的自變數即使「非常不顯著」,它仍然會與應變數產生某種程度的關聯(即使是隨機的)。因此,它總會解釋到一些變異。SSR 的計算是基於模型對數據的擬合程度。加入任何一個變數(無論顯著與否),只要它與應變數不是完全不相關(例如,隨機加入一個與應變數完全無關的變數,其 SSR 增加量會趨近於零,但通常不會是負的),都會在某種程度上「增加」模型對變異的解釋。
    因此,SSR 會增加。

  2. R² (Coefficient of Determination):
    R2=SSRSSTR^2 = \frac{SSR}{SST}。總平方和 (SST) 是固定的(除非我們改變應變數)。由於 SSR 會增加,所以 R2R^2 的值也會增加。
    即使加入的變數非常不顯著,R² 也總是會增加(或至少不減少)。R² 永遠不會因為加入變數而減少。

  3. F statistic:
    F 統計量用於檢定模型的整體顯著性。對於複迴歸,F 統計量通常計算為:
    F=MSRMSE=SSR/dfregSSE/dferrorF = \frac{MSR}{MSE} = \frac{SSR / df_{reg}}{SSE / df_{error}}
    其中,dfregdf_{reg} 是迴歸自由度,dferrordf_{error} 是誤差自由度。
    當我們加入一個新的自變數時:

    • dfregdf_{reg} 會增加 1。
    • SSESSE (Sum of Squared Errors) = SST−SSRSST - SSR。由於 SSR 增加,SSE 會減少。
🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 19 題

  1. After a couple of days, the consultant developed this regression model: y = β₀ + β₁x₁ + β₂x₂ + β₃x₁² + β₄x₂² + ε. If the consultant wish to test the significance of higher-order terms (x₁² and x₂²), which test would they use?
    A) overall F test
    B) partial F test
    C) Durbin-Watson test
    D) t test
    E) Cook's distance measure

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查迴歸模型中,檢定特定變數(在此是高階項)顯著性的方法。

給定的迴歸模型為:y=β0+β1x1+β2x2+β3x12+β4x22+ϵy = \beta_0 + \beta_1x_1 + \beta_2x_2 + \beta_3x_1^2 + \beta_4x_2^2 + \epsilon。
這個模型包含了線性項 (x1,x2x_1, x_2) 和高階(二次)項 (x12,x22x_1^2, x_2^2)。

顧問希望檢定高階項 (x12x_1^2 和 x22x_2^2) 的顯著性。這意味著要檢定 β3\beta_3 和 β4\beta_4 這兩個係數是否同時為零。
即檢定:
H0:β3=0H_0: \beta_3 = 0 且 β4=0\beta_4 = 0
H1H_1: β3\beta_3 和 β4\beta_4 不全為零。

這種檢定兩個或多個係數是否同時為零的檢定,通常使用 F 檢定。特別是當我們想檢定一個「子集」的變數(在這裡是 x12x_1^2 和 x22x_2^2)是否對模型有顯著貢獻時,我們使用偏 F 檢定 (partial F test)。

偏 F 檢定的基本思想是比較兩個模型:

  1. 縮減模型 (Reduced Model):假設 H0H_0 為真,即 β3=0\beta_3 = 0 且 β4=0\beta_4 = 0。模型變為 y=β0+β1x1+β2x2+ϵy = \beta_0 + \beta_1x_1 + \beta_2x_2 + \epsilon。
  2. 完整模型 (Full Model):即原始模型 y=β0+β1x1+β2x2+β3x12+β4x22+ϵy = \beta_0 + \beta_1x_1 + \beta_2x_2 + \beta_3x_1^2 + \beta_4x_2^2 + \epsilon。

偏 F 檢定比較這兩個模型的 SSR (或 SSE) 的差異,並與誤差的均方進行比較。
F=(SSRfull−SSRreduced)/dfadded(SSEfull)/dferror,fullF = \frac{(SSR_{full} - SSR_{reduced}) / df_{added}}{(SSE_{full}) / df_{error, full}}

🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

第 20 題

  1. Finally, the consultant would like to visualize some of the models they developed for the client. The graph of the prediction
    equation obtained from the model y = β₀ + β₁x₁ + β₂x₂ + ε is a(n)
    A) line
    B) parabola
    C) exponential curve
    D) ellipse
    E) plane

登入後即可作答並保存紀錄。

這一題的完整詳解

本題考查複迴歸模型的可視化。

給定的模型是:y=β0+β1x1+β2x2+ϵy = \beta_0 + \beta_1x_1 + \beta_2x_2 + \epsilon。
這是一個線性迴歸模型,其中應變數 yy 是兩個自變數 x1x_1 和 x2x_2 的線性組合,再加上一個誤差項 ϵ\epsilon。
在沒有誤差項的情況下,預測方程是:y^=β0+β1x1+β2x2\hat{y} = \beta_0 + \beta_1x_1 + \beta_2x_2。

  • 如果只有一個自變數 (y=β0+β1x1+ϵy = \beta_0 + \beta_1x_1 + \epsilon),則其圖形是一條直線 (line)。
  • 如果模型包含二次項,例如 y=β0+β1x1+β2x12+ϵy = \beta_0 + \beta_1x_1 + \beta_2x_1^2 + \epsilon,則圖形是拋物線 (parabola)。
🔒

後續完整解題步驟與【答案】

免費註冊,享三天全站完整詳解閱覽。

免費註冊

其他考古題