USING STATISTICS TO DESCRIBE A STUDY SAMPLE
STATISTICAL TECHNIQUE IN REVIEW
Most studies describe the subjects that comprise the study sample. This description of the sample is called the sample characteristics which may be presented in a table or the narrative of the article. The sample characteristics are often presented for each of the groups in a study (i.e. experimental and control groups). Descriptive statistics are used to generate sample characteristics, and the type of statistic used depends on the level of measurement of the demographic variables included in a study (Burns & Grove, 2007). For example, measuring gender produces nominal level data that can be described using frequencies, percentages, and mode. Measuring educational level usually produces ordinal data that can be described using frequencies, percentages, mode, median, and range. Obtaining each subject’s specific age is an example of ratio data that can be described using mean, range, and standard deviation. Interval and ratio data are analyzed with the same type of statistics and are usually referred to as interval/ratio level data in this text.
Source: Troy, N. W., & Dalgas-Pelish, P. (2003). The effectiveness of a self-care intervention for the management of postpartum fatigue. Applied Nursing Research, 16 (1), 38–45.
Troy and Dalgas-Pelish (2003) conducted a quasi-experimental study to determine the effectiveness of a self-care intervention (Tiredness Management Guide [TMG]) on postpartum fatigue. The study subjects included 68 primiparous mothers, who were randomly assigned to either the experimental group (32 subjects) or the control group (36 subjects) using a computer program. The results of the study indicated that the TMG was effective in reducing levels of morning postpartum fatigue from the 2nd to 4th weeks postpartum. These researchers recommend that “mothers need to be informed that they will probably experience postpartum fatigue and be taught to assess and manage this phenomenon” (Troy & Dalgas-Pelish, 2003, pp. 44-5).
Relevant Study Results
“A total of 80 women were initially enrolled [in the study] … twelve of these women dropped out of the study resulting in a final sample of 68.” (Troy & Dalgas-Pelish, 2003, p. 39). The researchers presented the characteristics of their sample in a table format for the experimental and control groups (see Table 1). The researchers found no significant differences between the control and experimental groups for any of the demographic or attribute variables.
TABLE 1 Sample Characteristics by Group
1. What demographic variables were included in this study?
2. Which of the demographic variables provided ordinal level data? Provide a rationale for your answer.
3. What level of measurement is the data for race?
4. What statistics were used to describe race in this study? Were these appropriate?
5. Could a mean be calculated on the race data? Provide a rationale for your answer.
6. Describe the race of both the experimental and control groups. What does this tell you about the population of this study?
7. What statistics were used to describe age in this study? Were these appropriate? Provide a rationale for your answer.
8. Were the groups similar in age? Provide a rationale for your answer.
9. What was the mode for the type of feeding provided by the experimental and the control groups? Is this mode what you would have expected?
10. Did the experimental group earn similar income to the control group? Provide a rationale for your answer.
ANSWERS TO STUDY QUESTIONS
1. Demographic variables described in the study were: age, income, length of labor, return to work, number of hours working per week, race, marital status, education, type of feeding, and amount of household and infant care responsibilities.
2. The variables education and amount of household and infant care responsibilities are both measured at the ordinal level since the data for each is sorted into categories that can be rank ordered. With education, high school is the lowest level of education, some college is the next level of education, and college graduate or higher is the highest level of education. Care responsibilities include ordinal data that are ranked from a low of “None” to a high of “All.”
3. The data collected for race is nominal level since race was measured using mutually exclusive categories of White, Black, Interracial, and Middle Eastern that cannot be rank ordered.
4. Frequencies and percentages were used to describe race for the experimental and control groups. Since the data are nominal, frequencies and percentages were appropriate. The researchers might have also identified the mode, which was White.
5. No, a mean cannot be calculated on the race data. A mean can only be calculated on interval and ratio level data that have numerically equal distances between intervals and not on nominal level data that can only be organized into categories. (See Exercises 1, 2, and 3, which are focused on identifying the level of measurements.)
6. Both the experimental and control groups are predominantly White, 92% and 96.55%, respectively. Thus, the sample is predominately White, and the results are reflective of a White or Caucasian population and not Black, Interracial, or Middle Eastern populations.
7. Age was described for both the experimental and control groups using means and standard deviations. The exact age of the subjects was obtained, providing ratio level data that are descriptively analyzed with means and standard deviations. The researchers might have also provided the range for age for both experimental and control groups.
8. The groups were very similar in age since the mean age for the experimental group was 26.72 and the mean age for the control group was 26.89. The distribution of the ages for the experimental and control groups were also very similar, with standard deviation of 5.05 for the experimental group and 5.25 for the control group.
9. Bottle-feeding was the mode for the experimental (53.1%) and the control (50%) groups since it was the most frequent type of feeding used by both groups. Either a “no” or “yes” answer is correct here as long as you provide a rationale. No, one might expect the mode to be breastfeeding since these were first-time mothers (primiparous) and breastfeeding has such positive outcomes for both infant and mother. Yes, one might expect bottle-feeding to be the mode since many of these mothers planned on returning to work.
10. No, the incomes were not similar for the two groups, but nor was the income significantly different for the groups. The means (M) and standard deviations (SD) for income indicate that the experimental (M = $35,675; SD = $23,969) and control groups (M = $41,450; SD = $17,527) were different. The control group subjects had an M, or mean, that was $5,775 higher than the experimental group, and the SD was much higher ($6,442) for the experimental group, indicating a larger range of incomes for that group. However, the narrative from the study indicated that the groups were not significantly different for any of the demographic variables.
Name:____________________________________________ Class: ____________________
? EXERCISE 11 Questions to be Graded
1. What demographic variables were measured at least at the interval level of measurement?
2. What statistics were used to describe the length of labor in this study? Were these appropriate?
3. What other statistic could have been used to describe the length of labor? Provide a rationale for your answer.
4. Were the distributions of scores similar for the experimental and control groups for the length of labor? Provide a rationale for your answer.
5. Were the experimental and control groups similar in their type of feeding? Provide a rationale for your answer.
6. What was the marital status mode for the subjects in the experimental and control groups? Provide both the frequency and percentage for the marital status mode for both groups.
7. Could a median be determined for the education data? If so, what would the median be for education for the experimental and the control groups? Provide a rationale for your answer.
8. Can the findings from this study be generalized to Black women? Provide a rationale for your answer.
9. If there were 32 subjects in the experimental group and 36 subjects in the control group, why is the income data only reported for 30 subjects in the experimental group and 34 subjects in the control group?
10. Was the sample for this study adequately described? Provide a rationale for your answer.
Grove, Susan K. Statistics for Health Care Research: A Practical Workbook. W.B. Saunders Company, 022007. VitalBook file.