Unit 6 - Practice Quiz

The null hypothesis () is a default statement that there is no relationship between two measured phenomena, or no association among groups. It is the hypothesis that is presumed true until statistical evidence indicates otherwise.

A Type I error, also known as a 'false positive', occurs when the null hypothesis is true, but we incorrectly reject it based on our sample data.

A Type II error, also known as a 'false negative', occurs when the null hypothesis is false, but we do not have sufficient evidence to reject it.

The symbol (alpha) represents the level of significance, which is the maximum acceptable probability of making a Type I error.

The Z-test is used when the population variance () is known. If the sample size is large (n>30), the sample variance can also be used as a good estimate, but the primary condition is known population variance.

The Z-test statistic is standardized so that it follows a standard normal distribution, which has a mean of 0 and a standard deviation of 1.

This test determines if there is a statistically significant difference between the means of two independent groups, typically when population variances are known or sample sizes are large.

The primary condition for using a Z-test is that the population standard deviation, , is known. If it is unknown, a t-test is the appropriate choice, especially for smaller samples.

The t-test is specifically designed for situations where the population variance is not known and has to be estimated from the sample data, which is common with smaller sample sizes.

With a large number of degrees of freedom (which corresponds to a large sample size), the t-distribution becomes nearly identical to the standard normal (Z) distribution.

Similar to the Z-test for difference of means, the t-test compares the averages (means) of two groups to see if they are statistically different from each other.

The shape of the t-distribution depends on the degrees of freedom (related to sample size). Therefore, to look up a critical value, you must specify both and the degrees of freedom.

The F-test is a statistical test used to assess whether two population variances are equal. It is also the fundamental test used in Analysis of Variance (ANOVA) to compare means by analyzing variances.

The F-statistic is defined as the ratio of two sample variances, . This ratio is used to test the hypothesis that the population variances are equal.

ANOVA uses the F-test to determine whether the variability between group means is larger than the variability of the observations within the groups.

The shape of the F-distribution is determined by the degrees of freedom for the variance in the numerator and the degrees of freedom for the variance in the denominator.

The goodness of fit test checks whether the observed frequency distribution of a categorical variable matches an expected or theoretical distribution (e.g., 'Are the dice fair?').

The Chi-square test is specifically designed for analyzing categorical data, where data is counted and placed into different categories.

The formula for the Chi-square statistic involves summing the squared differences between observed (O) and expected (E) frequencies, divided by the expected frequencies: .

A small value means there is a small difference between the observed and expected data, suggesting a good fit between the sample data and the hypothesized distribution. This would lead to not rejecting the null hypothesis.

A Type I error occurs when a true null hypothesis is rejected. In this case, the null hypothesis (drug has no effect) was true, but the researchers rejected it, leading to a false positive conclusion.

The Z-statistic is calculated using the formula: . Plugging in the values: .

The t-statistic is calculated as . Here, the null hypothesis is that the mean improvement is 0. So, .

The F-statistic is the ratio of the larger variance to the smaller variance, . The degrees of freedom are for the numerator and for the denominator.

The expected frequency for a category is calculated by multiplying the total sample size by the claimed proportion for that category. Expected frequency for A = .

The significance level is the probability of a Type I error, so decreasing it directly decreases the chance of a Type I error. However, making it harder to reject the null hypothesis (decreasing ) increases the chance of failing to reject a false null hypothesis, thus increasing the probability of a Type II error, .

The standard error of the difference is calculated as . Here, hours.

The calculated t-statistic (2.50) is greater than the critical value (1.833). Therefore, the result falls in the rejection region, and we reject the null hypothesis, concluding that the training program has a statistically significant effect.

The degrees of freedom are . Since the calculated statistic (12.0) is greater than the critical value (11.07), we reject the null hypothesis (). This means there is a significant difference between the observed and expected frequencies.

The F-test is specifically designed to test the null hypothesis that two normally distributed populations have equal variances (). It forms the basis for comparing variability between two groups.

The formula for pooled variance is . Plugging in the values: .

Power is the probability of making the correct decision when the null hypothesis is false. It is the ability of a test to detect a real effect. Since is the probability of a Type II error (failing to reject a false ), is the probability of avoiding that error.

The correct option follows directly from the given concept and definitions.

For a one-sample t-test, the degrees of freedom (df) are calculated as the sample size minus one. In this case, .

The contribution of one category to the Chi-square statistic is calculated as . For this category, the calculation is .

The F-test for variances is quite sensitive to departures from normality. Therefore, a critical assumption is that both underlying populations follow a normal distribution. Other assumptions include independence of samples.

The correct option follows directly from the given concept and definitions.

For an independent samples t-test with the assumption of equal variances, the degrees of freedom are calculated as . In this case, .

The primary purpose of the Chi-square goodness-of-fit test is to compare observed sample frequencies with the expected frequencies that would occur if the sample were drawn from a population with a specific theoretical distribution (e.g., uniform, normal, or a specific ratio).

The F-statistic is the ratio of two sample variances, . If this ratio is exactly 1.0, it means that the numerator and denominator are equal, so the sample variances ( and ) are identical. It does not, by itself, prove the population variances are identical.

A p-value of 0.08 with a small sample size trying to detect a small effect is a classic sign of an underpowered study. Power is the probability of correctly rejecting a false null hypothesis (). Failing to reject H0 doesn't prove it's true. Increasing the sample size () increases power, which means it decreases the probability of a Type II error (). The Type I error rate () is fixed by the researcher and is not affected by sample size.

A two-sided hypothesis test at level will reject if and only if the hypothesized value is outside the % confidence interval. For test (1), the value 102 is outside the 95% CI [98.5, 101.5], so we reject . For test (2), a one-sided test at has the same critical value as a two-sided test at . The sample mean is the center of the CI, . Since 98 is outside the CI and the sample mean is in the direction of (), we reject for the one-tailed test as well.

The pooled t-test is not robust to violations of the equal variance assumption when sample sizes are unequal. When the group with the smaller sample size also has the larger population variance, the pooled variance estimate is biased. This leads to an underestimation of the true standard error of the difference in means. A smaller denominator in the t-statistic results in a larger |t|, making it easier to reject the null hypothesis. Consequently, the actual Type I error rate becomes much higher than the specified .

An F-statistic close to 1 provides evidence in favor of the null hypothesis that the variances are equal, so we would fail to reject . However, a critical caveat of the F-test for equality of variances is its extreme sensitivity to the normality assumption. Even slight departures from normality in the underlying populations can severely compromise the test's validity, making it unreliable in many real-world applications. Therefore, while we don't reject , we must acknowledge this critical limitation.

The degrees of freedom for a Chi-square goodness-of-fit test are calculated as , where is the number of categories and is the number of parameters estimated from the data used to calculate the expected frequencies. Here, there are categories. The parameter for the Poisson distribution, , was not known beforehand but was estimated from the data using the sample mean. Therefore, . The degrees of freedom are .

The most severe consequence is a Type I error (approving a harmful drug). To minimize the risk of this error, the researcher should lower the significance level , which is the probability of a Type I error. This makes the criterion for rejecting the null hypothesis stricter. However, there is a trade-off: decreasing for a fixed sample size increases , the probability of a Type II error. The researcher is thus making a conscious choice to be more cautious about false positives at the expense of potentially missing a true effect.

With extremely large sample sizes, even trivial differences can become statistically significant. The p-value of 0.04 indicates statistical significance (if ), meaning we can be confident there is a non-zero difference between the population means. However, the magnitude of this difference (0.5) is very small compared to the variability within the groups (SD=50). This distinction between statistical significance (is there an effect?) and practical significance (is the effect large enough to matter?) is crucial, especially in 'big data' scenarios.

The t-test's validity relies on the assumption that the sampling distribution of the mean is approximately normal. The Central Limit Theorem ensures this for large sample sizes (often cited as ). However, with a small sample size like drawn from a heavily skewed population, the sampling distribution of the mean will also be skewed. This violates the core assumption of the t-test, making its p-value and confidence intervals unreliable. In this scenario, a non-parametric alternative like the Wilcoxon signed-rank test would be more appropriate.

This question tests the distinction between the F-test for equality of variances and the F-test within an ANOVA. The F-test in ANOVA compares the variance between groups to the variance within groups to make an inference about the population means. A fundamental assumption of ANOVA is homoscedasticity, meaning all groups are assumed to have equal population variances (). Rejecting the null hypothesis in an ANOVA says nothing about these variances; it only suggests that at least two population means are different. The equality of variances is a precondition for the test, not a result of it.

The Chi-square test's statistical theory relies on the test statistic following a Chi-square distribution, which is an approximation that holds well when expected frequencies are sufficiently large (a common rule of thumb is at least 5). When this condition is violated, the p-value can be inaccurate. The standard procedure is to collapse or pool categories in a logical manner until the expected frequency in each new, combined category meets the minimum requirement. This reduces the number of categories () and thus the degrees of freedom, but it makes the test valid.

This is the multiple comparisons problem. For a single test, the probability of NOT making a Type I error is . Since the 10 tests are independent, the probability of not making a Type I error in ANY of the 10 tests is . The event of 'at least one Type I error' is the complement of 'no Type I errors'. Therefore, the probability is , which is approximately 0.40. This demonstrates how the family-wise error rate inflates when conducting multiple tests.

A paired t-test properly accounts for the inter-subject variability by analyzing the differences within each subject. An independent t-test incorrectly treats this variability as random error. With a strong positive correlation between measurements, the variance of the differences (used in a paired test) is much smaller than the sum of the variances of the two groups (used in an independent test). By using the independent test, the researcher includes the large variability between subjects, which inflates the calculated standard error, reduces the t-statistic, and makes it much harder to detect a true effect (i.e., it decreases the power of the test).

The expected value of a Chi-square distribution is its degrees of freedom, which in this case is . A test statistic of 0.5 is extremely close to zero, indicating that the observed frequencies are much closer to the expected frequencies than one would expect due to random sampling variability. This results in a very high p-value (close to 1.0). While this means we do not reject the null hypothesis, a value this low can be suspicious. It suggests the data fits the model 'too well,' which can be a red flag for issues like data fabrication or a misunderstanding of the random process being studied.

The p-value for a two-tailed test with is . Since , we would fail to reject in the two-tailed test. However, for a one-tailed test with , the p-value is simply the area in the tail corresponding to the test statistic, provided the statistic is in the direction of the alternative. Since is negative, it is in the direction of . Therefore, the one-tailed p-value is . Since , we would reject in the one-tailed test. This shows how the same data can lead to different conclusions depending on the hypothesis.

The F-distribution is not symmetric. The critical values are denoted . A property of the F-distribution is that . The lower critical value for the first test (with df) is the reciprocal of the upper critical value for an F-distribution with the degrees of freedom swapped. Therefore, the upper critical value for the second test () is the reciprocal of the lower critical value for the first test ().

A confidence interval estimates the range for the population mean. A prediction interval estimates the range for a single future data point. The uncertainty in predicting a single point comes from two sources: the uncertainty in estimating the true population mean (captured by the confidence interval) and the inherent random variability of individual data points around the mean. Because it incorporates this additional source of variance, a prediction interval is always wider than a confidence interval for the mean calculated from the same data.

While the Student's t-test is technically the correct test when population variances are unknown, the t-distribution converges to the standard normal (Z) distribution as the degrees of freedom increase. With sample sizes of 40 and 50, the degrees of freedom will be large enough (df > 30) that the t-distribution is nearly identical to the Z-distribution. In this case, using the sample standard deviations () as plug-in estimates for the population standard deviations () in the Z-test formula is a common and acceptable approximation.

Power is the probability of correctly rejecting a false null hypothesis. It depends on the effect size, which is the difference between the true parameter value and the value under the null hypothesis. The original power calculation was based on an effect size corresponding to . If the true mean is , the true effect size is larger ( vs ). A larger effect size is easier to detect, meaning the distribution under the true alternative hypothesis is further from the null distribution. This increases the probability of rejecting the null hypothesis, thus increasing the power of the test to a value greater than 0.80.

The Welch-Satterthwaite equation for degrees of freedom often produces a non-integer value. When using a t-distribution table (or software that requires integer df), the standard conservative practice is to truncate the value (round down) to the nearest integer. In this case, using instead of or . The t-distribution has slightly thicker tails for lower degrees of freedom. By using the lower integer df, the calculated p-value will be slightly larger, making the test more conservative and reducing the risk of a Type I error.

The maximum value of the statistic occurs in the most extreme case where one category contains all observations and the other categories contain zero observations. Let's assume the -th category has all observations, so and for . The test statistic becomes . This sum simplifies to . This demonstrates that while the statistic can be very large, it is bounded by a function of the sample size and the number of categories.