Keywords

Cultural consensus analysis refers to “a collection of analytical techniques and models that can be used to estimate cultural beliefs and the degree to which individuals know or report those beliefs” (Romney et al., 1987, p. 339). The model “estimates each individual’s level of cultural competency (the degree to which each individual shares the group or normative values) and the answer to each question” (Weller et al., 1999, p. 723). In other words, the model estimates individual differences in cultural knowledge about a particular domain (such as diabetes) and allows for testing variation across respondents based on variables associated with subgroups of respondents (Romney et al., 1987).

The cultural consensus model is based on three main assumptions: (1) each participant answers questions about a cultural domain independently from other participants, (2) the questions asked of respondents belong to one specific cultural domain, and (3) that there is only one answer for each question. (Romney et al., 1987; Weller et al., 1999). This model has been used in many studies to understand people’s knowledge about diabetes among US Latinos (Weller et al., 1999), the classification of disease among Mestizo and indigenous Guatemalans (Romney et al., 1986), postpartum hemorrhage among mothers and midwives in Bangladesh (Hruschka et al., 2008), and the causes, symptoms, and treatments of AIDS, diabetes, the common cold, empacho, and mal de ojo in Guatemala, Puerto Rico, and Mexico (Weller & Baer, 2001).

In this chapter, I report the results of applying the consensus model to answer two questions: (1) Do Arab Americans have homogeneous beliefs about the causes, symptoms, and treatments for diabetes? (2) If not, then are their discernible differences in their knowledge of diabetes based on sex, age, generation, education level, country of origin, and language in which the survey is taken. To answer these questions, I followed the methods detailed in Chap. 4 to build a cultural consensus survey and collect data from 78 Arab American Muslim respondents in Dearborn, Michigan.

6.1 Cultural Consensus Results

The consensus analysis routine in UCINET (Borgatti et al., 2002) produces four outputs of data: (1) a participant-by-participant agreement matrix (the percentage of times each participant agrees with every other participant in his or her answers to the questions on the knowledge and beliefs about diabetes test); (2) the first and second eigenvalues from the factor analysis of the agreement matrix, the number of negative factor scores (competencies), and the ratio of the first to the second eigenvalue; (3) a competence score for each respondent; and (4) the answer key found for the cultural consensus survey, calculated from the responses of the survey participants. If there is one cultural belief regarding diabetes among Arab Americans in Dearborn, MI, then we expect the first eigenvalue to be much larger than the second, and we also expect to have no negative individual competence scores (Romney et al., 1986, 1987; Weller, 2007).

Table 6.1 shows the first- and second-factor eigenvalues and the number of negative competencies (that is, negative values on the first factor) found in the data (N = 78). The ratio of the first eigenvalue (22.220) to the second (2.939) is 7.561, and, as shown in Table 6.2, there are no negative individual competence scores. This indicates that the data fit the consensus model, and that Arab Americans in Dearborn share an overall cultural model of diabetes. Their shared answer key is shown in Table 6.3. The shared answer key represents the overall shared model of diabetes among Arab American Muslims in Dearborn, Michigan. In general, the beliefs presented in the model, dose not contradict with the Western medical knowledge. One issue related to fasting in Ramadan can be noticed in the model. Arab American Muslims share the beliefs that people who have diabetes should fast in Ramadan (even if they are old and have health complications) and maintain their usual year-round diet while fasting.

Table 6.1 First and second factors’ eigenvalue, ratio, and number of negative competences
Full size table
Table 6.2 Individual competency scores
Full size table
Table 6.3 Diabetes overall knowledge answer key
Full size table

Although there is a single cultural model of diabetes among these Arab Americans in Dearborn, the 2nd-factor eigenvalue of 2.9 shows that there is possible variation in the model. To check whether there was enough variation to be tested in the model, I followed the recommendation of Romney et al. (1987) and checked the standard deviation for the respondents’ competence scores. The standard deviation for the average score is 0.21, which is above the 0.18 cutoff suggested by Romney et al. (1987), indicating sufficient variation to be tested in the model.

6.2 Testing Variation Using 2nd-Factor Loading

Using the loadings on the 2nd factor to look for substantive variation in a consensus model was suggested by Boster (1986), who posited the presence of systematic deviations from consensus “by pairs of informants who agree with each other more than would be expected on the basis of their approach to the general consensus” (Boster, 1986, p. 431). Boster (1986) referred to “agreement beyond the overall cultural consensus” as ‘‘residual agreement’’ (Dressler et al., 2015, p. 24). In their work comparing novice and professional fishermen, Boster and Johnson (1989) presented a method for examining residual agreements in a consensus model, and Dressler et al. (2015) describe the method for what they call “finding culture change in the second factor.”

To look for possible subgroups in the 2nd-factor loadings shown in Table 6.4, I correlated the 2nd-factor loading with the following variables: sex, generation (first or second generation), education, age, country of origin, and the language the survey was taken in (For statistical tests, SPSS software was used).

Table 6.4 Second-factor loading (N = 78)
Full size table

Eta coefficient was used to examine the correlation of the categorical variables with the 2nd-factor loadings. The Eta coefficient ranges from 0 to 1, with a score of 0.3 or greater considered sufficient to establish a significant correlation (Garson, 2008). As shown in Table 6.5, only one of the variables examined (generation; eta = 0.369) was significantly correlated with the 2nd-factor loading. We conclude, then, that there may be significant differences in cultural beliefs about diabetes between first- and second-generation Arab Americans. The correlation between the 2nd-factor loading and age was 0.52 (Table 6.6), indicating possible differences in cultural beliefs about diabetes based on age.

Table 6.5 Eta results for the categorical variables (N = 78)
Full size table
Table 6.6 Pearson’s correlation for age and second-factor loading (N = 78)
Full size table

6.3 Cultural Consensus Analysis Based on Generation and Age

In sum, there are two variables that can be used to check for possible subgroups in the cultural beliefs about diabetes among Arab Americans in Dearborn, MI: generation and education. Therefore, I first divided the sample based on generation into two subgroups, first generation (n = 40) and second generation (n = 38), and I ran the cultural consensus for each subgroup separately, again using UCINET. Table 6.7 shows the results. Both first- and second-generation cultural consensus results show that the 1st factor is much larger than the 2nd factor, and that there are no negative individual competence scores. This indicates that the data fit the consensus model, and that first- and second-generation Arab Americans in Dearborn, MI share a cultural model of diabetes.

Table 6.7 Results of culture consensus for first (N = 40) and second generation (N = 38)
Full size table

To further examine the differences between the first- and second-generation knowledge, I compared the answers keys for each group. Two differences were found between the answer keys for first-and second-generation subgroups: (1) For first-generation respondents, the consensus answer to the statement “People who have diabetes have the same risk of having depression as people who don’t have diabetes” was true, while for the second generation it was false. (2) For first-generation respondents, the consensus answer to the statement “People who have diabetes should be prevented from eating sugar at all” was true, while for the second generation it was false. These differences may be explained by the level of education—those born in the United States (the second generation) having higher levels of education, on average, than those of the first generation. It is also very important to think about the culture that drives those differences especially when it comes to mental illness in general and depression in specific. During my daily interactions with the community members, I noticed that first-generation Arab Americans tend to dismiss the idea of having mental illnesses, as they see it as a sign of “not having faith.” When chatting with one of the second-generation participants, she mentioned that the word “depression” is banned in their house by her parents (first generation). As she explained further that her parents believe that not praying and not staying close to God is the reason for feeling depressed.

To test for possible differences by age, the data were divided into two age subgroups: 18–24 (n = 57) and 25–53 (n = 21) and ran the consensus analysis separately on each subgroup. Table 6.8 shows the results. For both age groups, the 1st factor is much larger than the 2nd, and there are no negative individual competence scores in either group. This indicates that the data for both age subgroups separately fit the consensus model.

Table 6.8 Consensus analysis results for both ages group 18–24 (N = 57), and age 25–53 (N = 21)
Full size table

Next, I compared the answer key for the age group 18–24 with the answer key for age group 25–53. Four differences were found: (1) Younger respondents (18–24) answered the statement “lack of exercise can cause diabetes” as true, while older respondents (25–53) answered it as false. (2) Age subgroup 18–24 answered the statement “people who have diabetes have the same risk of having depression as people who don’t have diabetes” as true, while those 25–53 answered it as false. (3) Age subgroup 18–24 answered the statement “people who have diabetes can fast during Ramadan” as true, while those 25–53 answered it as false. (4) Age subgroup 18–24 answered the statement “people who have complications from diabetes can still fast during Ramadan” as true, while the age subgroup 25–53 answered it as false.

Regarding the differences in answers to the first statement, people with higher education levels (ages 18–24) are more likely to have higher awareness of the importance of exercise on health. For statements 2, 3, and 4, it’s possible that people in age subgroup 25–53 have more health complications, making them aware of health and depression as well as the possibility of not fasting during Ramadan.n