EXPLANATION OF TABLE: This table is a fundamental summary of this chapter and a link with the following chapter. It shows the regression results of each of seven y variables (denoted 'SYMPTOMS') on to the nine x variables indicated as culturally universal predictors by best subsets regression analysis: from 6 daily habits, level of alcohol consumption, change in alcohol consumption, frequency of exercise; from 14 sources of social support, from a spouse, from a close friend; from life events, contemplation of counselling, action on counseling, age, parenting status. The key statistic is the R' statistic, which indicates the variance explained by the predictor variables. The bottom of the table indicates p values for x variables which are significant; these are useful as indicators of cultural 'specifics'.
SXMPTOM CULTURAL DIFFERENCES IN R», REGRESSION
UK JAPAN HONG KONG
Intrusion R*=10.9 R'=7.5 R*=30.0 (F = 0.93, p = 0.496) (F = 1.55, p = 0.134)a (F = 1.26, p = 0.268} Avoidance R* =529.1 R'=9.4 R»=18.2 (F = 3.13, p = 0.005) (F = 1,96, p = 0.047)b (F = 1.21, p = 0.310) Anxiety R*=7.7 R*«9.0 R*=45.5% (F = 0.62, p = 0.757) (F = 1.88, p = 0.059)c (F = 4.44, p = 0.000)§ Depression R*s9.2 R*=slG.9 R'=27.0 (F = 0.76, p = 0.636) (F = 2.30, p = 0.01B)d (F = 2.01, p = 0.058)t Perf Diff R%=10.4 R*«10.2 R»=30.2 (F = 0.87, p=0,557)‘ (F = 2.13, p = 0.030)e (F = 2.17, p = 0.043)t Distress R*=6.7 R*=10.6 R*=25.6 (F = 0.54, p = 0.822) (F = 2.23, p = 0.022)f (F = 1.83, p = 0.086)# Souiatics R*=8.9 R»=10.7 R'=28.3 (F = 0.70, p = 0.690) (F = 2.27, p = 0.020)g (F = 2.02, p = 0.058)#
Key to X variables which are significant: FOR UK
Alcohol change (0.003) FOR JAPAN
a Contemplation of counselling (0.027) b Contemplation of counselling (0.01)
c Support from close friend (0.037), parenting status (0.05) d Sport interest (0.001)
e Contemplation of counselling (0.009), age (0.047) f Contemplation of counselling (0.003)
g Support from partner (0.034), age (0.019) FOR HONG KONG
§ Alcohol level (0.014), Alcohol change (p.024) support from close friend (0.01) t Support from work colleague (0.027)
+ Contemplation of counselling (0.009), support from close friend (0.016) # Support from close friend (0.05)
## Contemplating counselling (0.049)
Details for v variables; Impact of Event Scale, Intrusion and Avoidance factors; Hospital Anxiety and Depression Scale, Anxiety and Depression factors; Hopkins Symptom Check List, Performance Difficulties, Somatic Distress and General Feelings of Distress.
ENDNOTES
’ To further understand the magnitude of differences, consider percentages of the scale maximum. Over all seven symptoms, mean level for the UK is 21.60%. For Japan mean level is 27,77%. For Hong Kong mean level is 43.80%. Thus, in broad terras Hong Kong symptomatology in general pathology and specific trauma reactions are, in effect, twice levels in UK, and 16.50% higher than in Japan.
^ Calculation of significant difference: The test for significant difference used in this chapter is based on the t-distribution tests described, inter alia, by Sprent (1977) and tabulated, inter alia, in Kmietowicz & Yannoulis (1975), although this test is in common usage throughout the social sciences. In short, tests of difference are calculated by a division of numerator by denominator. This is undertaken such that an arising value (a 't- value') can be compared to the one-tailed tables from which levels of significance can be ascertained at varying degrees of freedom (denoted v and varying from l to 300, then infinity) . In the t-test for significant difference, the numerator is the difference between the mean values of the variables observed in the populations under investigation. The denominator is the square root of the sum of the squares of the standard deviations divided by the sample n. in both populations.
An example from Table XII follows. To test for difference between UK and Japan intrusion symptoms, the numerator is 2.67 (mean score for Japan) less 2.31 (mean score for UK). This = 0.36. The denominator is the square root of the sum of the following: first, the square of 2.93 (UK standard deviation in intrusion symptoms) divided by 68 (the UK sample number); second, the square of 3.74 (Japan standard deviation in intrusion symptoms) divided by 688 (the Japan sample number). In arithmetic form:
(10.28)-(12.27)
V[(2.93)^ 4- 668] + [(3.74)^
h- 688 ]
This produces a t-value of approximately 5.10. Given this substantially exceeds the critical points at 50 degrees of freedom (t=3.4602) this suggests a significant difference between Japan and UK in intrusion symptoms at p<0.0005. Critical points at 60 degrees of freedom - which is chosen because it is the most conservative or lowest sample - are as follows; p<0.01, t=2.3901; p<0.005, t=2.6603; p<0.001, t=3.2317; p<0.0005, t=3.4602 (Kmietowicz & Yannoulis, 1976; 21; further details can be found in most statistical texts such as Kanji (1993); Sprent (1977)).
^ Levels of significance as follows (Kmietowicz & Yamnoulis, 1976: 24): UK: n.=C.70; p<0.01, r=0.2737; p<0.00S, r=0.3017; p<0.001, r=0.3583. Japan: n.=C.500; p<0.01, r=0.1038; p<0.005, r=0.1149; p <0.001, r=0.1376.
Hong Kong: n.=c.60; p<0.01, r=0.2948; p<0.005, r=0.3248; p<0.001, r=0.3850.
* " [Best subsets regression] can be used to select a group of 'best subsets' for further analysis. The general method is to select the smallest subset that fulfills certain statistical criteria. The motivation for variable selection is based on the fact that the subset model may actually estimate the regression coefficients and predict future responses with smaller variance than the full model using all m predictors" (Minitab, 1987: 84)
® "In general, we look for models where Cp [c-p] is small and is also close top. If the model is adequate (i.e., fits the data well), then the expected value of Cp is approximately equal to p, the number of parameters in the model. A small value of Cp indicates that the model is relatively precise (has small variance) in estimating the true regression coefficients and predicting future responses. This precision will not improve much by adding more predictors. Models with considerable lack of fit have values of Cp larger than p." (Minitab, 1987: 82)