统计问题第65问：描述数据的离散趋势

Question

Last week’s question discussed sample standard deviation. For a variable such as body mass index (BMI), the sample standard deviation provides a measure of the average difference between the BMI of individuals in the sample and the sample mean.

The example used last week was a cohort study: researchers investigated the association between slow walking speed and risk of death in older people.2 A cohort of 3208 French men and women aged over 65 years living in the community were recruited in 1999 and followed for an average of 5.1 years.

At baseline the cohort had a mean BMI of 25.6 kg/㎡with a standard deviation of 4.0 kg/㎡. The range in BMI that was no further than two sample standard deviations either side of the sample mean—that is, [25.6 – 2(4.0)] to [25.6 + 2(4.0)] kg/㎡—equalled 17.6 to 33.6 kg/㎡.

Which one of the following best describes the approximate percentage of individuals whose BMI at baseline was encompassed by this range?

a) 95% of the population

b) 99% or more of the cohort

c) 95% or more of the cohort

d) 68% of the population

Answer

Answer c—At baseline, approximately 95% or more of the cohort had a BMI that was no further than two sample standard deviations either side of the sample mean. The sample mean was 25.6 kg/㎡ and sample standard deviation 4.0 kg/㎡. Therefore, approximately 95% or more of the cohort had a BMI between [25.6 – 2(4.0)] and [25.6 + 2(4.0)] kg m-2—that is, between 17.6 and 33.6 kg/㎡.

The sample standard deviation can be used to calculate further ranges in BMI that contain certain percentages of the cohort members at baseline. For example, ranges in BMI that include approximately 68% or more of the cohort, or approximately 99% or more, may be derived. These ranges can be used to establish the potential spread in BMI for different quantities of the sample; however, it is the range that encompasses approximately 95% or more of the sample that is typically considered. Establishing the potential variation in BMI, plus other characteristics of the sample at baseline such as age, informs the extent to which the study results can be generalised and extrapolated.

At baseline, approximately 68% or more of the cohort had a BMI that was no further than one sample standard deviation either side of the sample mean. For the above study, this range equals [25.6 – 4.0] to [25.6 + 4.0] kg/㎡—that is 21.6 to 29.6 kg/㎡. Furthermore, at baseline approximately 99% or more of the cohort had a BMI that was no further than three sample standard deviations either side of the sample mean (answer b is false). This BMI range equals [25.6–3(4.0)] to [25.6+3(4.0)] kg/㎡—that is, 13.6 to 37.6 kg/㎡.

We derived the above ranges from the sample mean and standard deviation, facilitating description of the distribution of BMI for members of the cohort at baseline. However, we cannot use these ranges to describe the distribution of BMI in the population (a and d are both false).

These ranges can always be calculated regardless of the variable measured in the sample, providing the variable is on a continuous scale. Nonetheless, the derived ranges may incorporate values not originally observed in the sample.

所以答案是选择 c

每天学习一点，你会更强大！