What these results suggest is that higher preference for alternative health care coincides with healthier eating habits (self-assessed, question 15 and 18), which comes as no major surprise, but also for more, but more restless sleeping, which coincides with higher self-reported stress (Q22) and less ability to concentrate when sleep is interrupted (Q21), especially for women. These results suggest further investigation comparing sleeping and eating habits in this population, perhaps. Question 17 suggests for another example that restaurant marketers may want to investigate advertising more to traditional, rather than alternative health consumers.
¶ … consumption of alternative, as opposed to traditional, health care consumption, and describing different groups within that preference. A 37-question survey was composed and checked for apparent face validity by the research group, who collected 320 surveys around southeastern Pennsylvania through a convenience sample by offering a chance to win pizza coupons donated by an employer. Twenty-nine of the surveys were rejected as incomplete leaving a sample of 291 complete surveys or surveys missing only one answer. The most significant problem discovered was that many of the questions employed subjective descriptors like "satisfied" or "exercising" evoked questions of measurement or definition. Some of these questions were rejected but others were retained qualified that those terms were self-defined. For example, some respondents telecommute, so answered very low on "hours of work outside the house" even while holding down a full-time job. These types of confounds indicate why deploying large-scale surveys like this carry high cost testing for face and content validity because of the extensive labor collecting and analyzing results.
Analyses
One of the other lessons learned was that asking 37 questions on a survey results in a very lengthy report. While the data is valuable probing for similarities and differences within and between groups, representing all the data comprehensively becomes problematic in limited space so we describe select variables here that pertain most to our question regarding the profile for alternative health consumers, however they defined that to themselves. Overall, the response to our survey, either from selection bias intentional or accidental on part of administrators, self-selection by respondents, or other unknown affects, resulted in more females than males, as described in Table 1. Table 2 shows frequency distribution for overall mean response by age, which was continuous data that researchers classed into deciles for presentation.
Table 1. Question 1 (Q1)
Response by Gender
n
0 (Male)
1 (Female)
The higher representation of lower age classes demonstrate potential selection bias either from survey administrators or respondents, or unknown factors affecting the convenience sample like choice of location or willingness to fill out surveys. Younger people could have more disposable time, for example, or more desire for potential reward. The solution would be to take a scientifically random sample from all age groups, but if this age was the local demographic, then these results would describe that population.
Table 2. Question 2
Response by age
Age
n
16-25
26-35
36-45
31
46-55
33
56-65
11
66-75
6
76-85
2
n
Figure 1 presents results from Table 2 in a bar graph and demonstrates proportions between the various age cohorts. The implication is that later questions describe these age and gender demographics, and if results are phrased in those terms, are useful for generalizing to that specific population.
Table 3 below displays frequency distribution by educational attainment, and measures of central tendency for these first three questions are reported in Table 4.
Table 3.
Question 3: Response by education
1
22
1=some high school
2
63
2=high school diploma
3
79
3=some college
4
40
4=certif.or assoc. degree
5
81
5=bachelors or masters
6
6
6=PhD or professional degree n =
Figure 2 below describes employment outside the house in hours. This information is displayed in a pie chart instead of bar graph or histogram because of the wide variation across means, where placing 39% and 1% on the Cartesian plane creates so small a distance between y-axis tick marks that the lowest mean entry becomes barely visible, and thus the utility of display using those modes is negligible. Frequency distribution does not suffer this drawback but other questions were more important but figure 2 helps to describe general sample characteristics and thus describe the subjects we focus more attention on below. Most of the respondents were working people, with the majority working nearly full or full time.
Table 4 presents the highest number of hours per week spent exercising. This table displays measures of central tendency that change when we delete especially high values that have low frequency. The range falls considerably when the highest single value is deleted, which indicates there was only one; likewise the mean continues to fall as we exclude the extreme values which may be potential outliers. Given that the median and mode do not change, the potential these values are outliers rather than included in the data set introduces skew, suggested by the change in means approaching the central value or median. Zero is not an outlier because as the mode, this must be the most frequent value. These dynamics suggest sample means may indicate questionable results without further testing to verify if skew distorts generalizability. There are plausible cases where such an extreme value would be kept, say if we were testing for risk of a rare event, in which case that would be the data point of interest compared to the rest.
Table 4. Partial Data Sample
from Q6, hours exercising
25
25
25
25
25
25
30
30
30
30
30
30
40
40
40
40
80
n =
SUM
mean
5.04
4.78
4.54
median
3.00
3.00
3.00
mode
0
0
0
range
80
40
30
To investigate potential variables of interest, the data was sorted into male and female, and health-care related questions of interest displayed for both groups. This convenient procedure shows that while both groups scored similarly on many questions, the difference between means for questions 20-23 may yield interesting results with further study.
Comparing means between groups this way often delivers interesting results. Considering means for exercise rates by hour for different education levels (Figure 4), does not reveal any particular trend.
Displaying mean hours worked by education does however suggest further research may reveal correlation between education level and hours of work, which could then be tested for likelihood at various levels of accuracy.
Displaying smoking incidence by education level reveals a clear trend that suggests higher education may coincide with better health, if smoking coincides with negative health outcomes.
Sorting the data by Likert-scale means ranking for preference for "alternative" health care, however the respondent defined that deliberately ambiguous and subjective term, resulted in a continuous ranking of ten classes. The rest of the means for various questions can then be compared against each other within these groups. Figures 7 and 8 display various means for the highest and lowest rankings for preference for alternative health care, means of one and ten respectively. Figure 7 shows data with higher entries in order to avoid scaling low answers so small as to impede usefulness of display.
In this sample at least, higher-educated respondents who rated alternative health care the highest had higher mean education than those who were least likely to prefer that alternative (Fig. 7, Q3). Interestingly, higher mean preference for alternative health coincided with mean fewer hours exercising, and not surprisingly, less smoking.
Quick display of other variable means via bar graph shows some will be more interesting than others given limited resources of time and report space. Reducing the scale on the y-axis displays an apparently wider difference on Q6 than in Fig. 7, which illustrates the value of comparing similar results on appropriate scales.
Table 5. Lowest vs. highest answers on Q23 "alt health"
Answered "1" on "prefer alt. health"
Q1
Q2
Q3
Q20
Q21
Q22
Q23
n (Males)
38.00
38.00
38.00
38.00
38.00
38.00
38.00
n (Females)
37.00
37.00
37.00
37.00
36.00
37.00
37.00
Mean Male
0.00
31.74
3.21
4.00
4.39
4.76
1.00
Mean Female
1.00
30.97
3.27
4.92
5.50
5.51
1.00
Subample Mean
0.49
31.36
3.24
4.45
4.93
5.13
1.00
Variance (M)
0.00
2.33
9.24
11.38
8.83
0.00
Variance (F)
0.00
1.20
12.41
10.31
10.70
0.00
Sample STDEV
0.50
12.45
1.32
3.30
3.32
3.13
0.00
STDEV (M)
0.00
14.50
1.53
3.04
3.37
2.97
0.00
STDEV (F)
0.00
10.11
1.10
3.52
3.21
3.27
0.00
Answered "10" on "prefer alt. health" n (Males)
7.00
7.00
7.00
7.00
7.00
7.00
7.00
n (Females)
14.00
13.00
14.00
14.00
14.00
14.00
14.00
Mean Male
0.00
37.14
3.43
5.86
5.57
5.71
10.00
Mean Female
1.00
39.69
3.50
5.64
6.79
6.21
10.00
Subample Mean
0.67
38.80
3.48
5.71
6.38
6.05
10.00
Variance (M)
0.00
93.14
1.95
11.81
11.95
13.90
0.00
Variance (F)
0.00
2.73
9.79
9.26
7.72
0.00
Sample STDEV
0.48
12.75
1.54
3.15
3.15
3.04
0.00
STDEV (M)
0.00
9.65
1.40
3.44
3.46
3.73
0.00
STDEV (F)
0.00
14.44
1.65
3.13
3.04
2.78
0.00
You’re 80% through this paper. Sign up to read the full paper.
Sign Up Now — Instant Access Already a member? Log inAlways verify citation format against your institution’s current style guide requirements.