help button home button ClinMed NetPrints
HOME HELP FEEDBACK BROWSE ARTICLES BROWSE BY AUTHOR
Warning: This article has not yet been accepted for publication by a peer reviewed journal. It is presented here mainly for the benefit of fellow researchers. Casual readers should not act on its findings, and journalists should be wary of reporting them.

This Article
Right arrow Abstract Freely available
Services
Right arrow Similar articles in this netprints
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Shahtahmasebi, S.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Shahtahmasebi, S.
Related Collections
Right arrow Paediatrics:
Other Paediatrics

Right arrow Public Health:
Smoking

clinmed/2003030003v1 (April 1, 2003)
Contact author(s) for copyright information

The correlates of teenage smoking: some problems with interpreting the evidence

Running title: The correlates of teenage smoking

by

Said Shahtahmasebi, Ph.D

Senior Lecturer/Research Methodology Consultant

School of Mathematics & Statistics

Faculty of Health & Sciences

CPIT

P.O. Box 540

Christchurch.

Tel: +64 3 940 8191 Fax: +64 3 940 8622

Email: saids@cpit.ac.nz

 

 

 

An earlier version of this paper was presented at the Australian Epidemiological Association, "Epidemiology in a globalised world", 11th Annual Scientific Meeting, 5-6 September 2002, Wellington, New Zealand.

February 2003.

The correlates of teenage smoking: some problems with interpreting the evidence

Summary

Background: This paper attempts to address the methodological problem of disentangling complex interrelationships between teenage smoking and other variables when using survey data. As in any research based on survey data, it is difficult to distinguish systematic from random patterns due to other variables.

Method: A pragmatic approach of classifying variables into three groups of objectively measured demographic, socio-environmental variables and subjectively measured socio-psychological variables was adopted. A secondary cross-sectional survey data source was used to fit and test three models.

Results: The results suggest socio-environmental and socio-psychological variables may effect teenage smoking through demographic and some other variables not measured in the survey.

Conclusion: Therefore, it is possible that the effects of variables reported in the literature on teenage smoking, if true, are over estimated. For example, self-esteem may not be as important and the role of peer pressure may not be as clear it has been made out to be.

 

Key words: systematic effect, subjective variables, objective variables

The correlates of teenage smoking: some problems with interpreting the evidence

1-Introduction

Most studies of teenage smoking have relied on observational or questionnaire data. As in any research based on survey data, it is difficult to distinguish systematic effects from random variation due to other variables. One of the fundamental methodological problem is disentangling the complex inter-relationships between smoking and other variables. Shahtahmasebi and colleagues; have addressed some of the difficulties associated with the analysis of survey data and reported an improved cross-sectional multivariate analysis to incorporate two fundamental issues of multicollinearity and the direction of causality. Neither of these problems is readily addressed in the smoking literature; control in observational studies is always problematic because of poorly measured, unmeasured or possibly unsuspected control variables, while methods for identifying direction of causality are very demanding of data and tend not to be robust to violations of statistical assumptions. Most surveys include variables that are subjectively measured e.g. self-reported health, perceptions and attitudes, and worrying about things. The inclusion of subjective variables exacerbate the problem twofold; (i) because of measurement error associated with the subjective variables, and, (ii) by adding confusion to the direction of causality. For example, the relationship between high self-esteem and smoking could be working in either direction. The consequences of including such variables in the analysis could lead to an over inflated relationship with the outcome variable and a possible mis-specification error. In this paper a secondary data set is used to demonstrate complex interactions and multicollinearity that is inherent to survey data and that it would be unwise to make inference about the data from cross-sectional analysis without exercising caution.

Background

A large number of variables have been reported to be associated with teenage smoking including demographic, social, environmental, economic, emotional and psychological variables;;;;;;. Some authors appear to concentrate on, and, distinguish between aspects of human behaviour as independent effects on smoking e.g. the psycho-social effect of wishing to belong to a peer social group, self-esteem and parental smoking behaviour, psychological process of risk assessment and risk taking;;. Others suggest smoking behaviour is likely to be influenced through the family environment, which affects psychological well-being, adjustment and problem behaviour, perceived academic performance and school conduct;. This ambiguity in interpreting the results is further exacerbated by the subjective effect of smoking e.g. feelings of higher self-esteem and being in control;; a perceived 'benefit' effect as utilised in the risk models;. The reported relationship between high self-esteem and smoking serves to demonstrate these difficulties. Quite apart from the problems associated with measurement of self-esteem and multicollinearity, the main issue with such variables is change over time. To infer from cross-sectional data a relationship between smoking and high self-esteem without the knowledge of levels of self-esteem prior to smoking is unwise, and, may lead to erroneous conclusions. For example, Murphy et al report low self-esteem among the female group who indicated an intention to smoke. This means that the association between higher self-esteem and smoking may become untenable, as it is likely that the individuals had lower self-esteem prior to taking up smoking. Therefore, it is not surprising that Paavola et al applying a cross-sectional statistical technique to longitudinal data, report a high correlation between smoking status in adolescence and in adulthood - as a cessation will often be accompanied by feelings of unease and anxiety;. Yet, survey studies often treat a mix of observations (e.g. parental smoking behaviour and self-esteem, or, psychological and environmental measures) as independent correlates (i.e. that no other effects are present) of smoking. Such an approach ignores the consequences of the multicollinearity in data.

I do not intend to provide a literature review on smoking behaviour, the interested reader is referred to Stead;;;. This study is concerned with an examination of the evidence from a cross-sectional analysis of the relationship between social characteristics and teenage smoking. The analysis seeks to identify any effect of social circumstances over and above the better understood effects of age and sex. The statistical modelling approach described seeks to achieve this objective by identifying the systematic relationships between the explanatory variables and smoking, distinguishing these relationships from irregular, random variations which tend to obscure any patterns in data on the one hand, and the misleading effects of multicollinearity on the other. Social circumstances are broadly defined and cover behavioural, attitudinal, emotional and self-esteem (see Appendix I).

2-Methods

Data

The data for this study came from a sample survey of the population of under 16 year olds in Yorkshire and Humberside, UK. This data set was the result of a programme by Yorkshire Regional Health Authority to collect base-line and follow-up data on the population's life-style. The programme had the backing of the Local Education Authority (LEA) but was discontinued following the abolition of the Regional Health Authorities in the UK during 1994/95.

Around 60 secondary schools in the former Yorkshire Regional Health Authority geographical boundary, agreed to take part in the survey. Years 9 and 11 (age range 11-16) in these schools were surveyed in 1991 using a health related behaviour questionnaire. The survey was self-administered in the classrooms during a normal lesson session. For complete anonymity; names, addresses and area postcodes were not collected. Furthermore, pupils were required to place the completed questionnaire in an unmarked envelope and drop it into a post bag themselves. The post bag was then returned to the survey administrators.

The questionnaire covered topics related to pupils' attitudes and behaviour with regard to health e.g physical exercise and out of school activities, nutrition, social contacts, dealing with problems, attitudes to and the use of drugs (including smoking and drinking). Relevant variables which are reported in the literature including self-esteem, worrying about a number of issues e.g. family, friends and school, were extracted from the data set and are shown in Table 1. The outcome variable "smoking" is a dichotomous dummy variable and takes the value "0" if a non-smoker (those who claimed they have never smoked or have given up smoking) and the value "1" if a smoker (those who claimed they smoked either regularly/occasionally), see appendix I & II. Furthermore, the age variable was categorised according to the school year (year 9: age group 12/13; year 11: age group 14/15), mainly because the cell frequencies for the lower and upper tails of the age range were too small to be meaningful.

Table 1 - approximately here

Just over 10,000 pupils responded to the survey. Out of these there were 9230 cases with a valid response on the smoking variable. The preliminary results suggest roughly equal number of boys and girls in the total sample and quite close to the ratio in the Regional population: 18% of the 11-16 year old smoke either occasionally or regularly; 60% of the smokers are girls and 40% are boys; around 22% of the girls and 14% of the boys are smokers.

Statistical analysis

A logistic regression model (see Appendix II) was fitted to the data with smoking variable (smoker/non-smoker) as the response variable. For model fitting the computer software package GLIM was used. The method of "forward substitution" successive model fitting was adopted with the "best" additional variable added to the model at each stage. The improvement in the model as a result of adding in each variable in turn was assessed by the likelihood ratio test statistic and the process ceased when no additional variable was significant at 5% level. This approach can be very instructive in an exploratory study. In particular, a marked change in the significance of a remaining variable when a new variable is added to the model is indicative of a spurious relationship arising from statistical association with the added variable. On the other hand, with high multicollinearity in the data, the final model may be heavily dependent upon a marginal choice between two variables at some stage in the model fitting process. It is important to detect when this is occurring to prevent undue emphasis being given to one variable where others are practically indistinguishable in their explanatory power. Moreover, identification of nearly interchangeable variables may guide further research in reducing multicollinearity problems by combining variables in composite indices.

As noted in the introduction, a problem arises, however, over the inclusion of the subjectively measured social and emotional factors (arbitrarily headed socio-environmental and socio-psychological respectively) in the analysis. Social circumstances will have an impact on the prevalence of teenage smoking, in part, by affecting emotional variables e.g. the desire to belong to a peer group. Controlling for these variables may therefore result in an over conservative, attenuated estimate of the effects of social circumstances. Reverse causality is also possible, with emotional circumstances influencing social circumstances. Excluding these variables may therefore exaggerate the effect of social circumstances. As emphasized earlier, we cannot unravel completely the complex causal relationships likely between social circumstances, emotional, and teenage smoking. However, the analysis could be improved to get an idea of the role of socio-environmental and socio-psychological variables as independent effects on smoking or as intervening variables between the objective (arbitrarily headed socio-demographic) variables and smoking. We can assess the role of subjective variables by forcing the direction of causality from objective variables to the outcome (smoking). Therefore, models were fitted with and without the socio-environment and socio-psychological variables;;. Three models were fitted to the data; firstly a model of objective variables was constructed based on the demographic variables. Secondly, the social variables were introduced to this model, and, thirdly the socio-psychological variables were then added to the second model.

3-Results

The role the different variables play in smoking can be examined by comparing the results from the three models in Table 2. The inclusion of socio-environmental factors (model 2) has a major impact on model 1 (Table 2): there are significant changes in the parameter estimates of the variables "gender", "age", "which parent live with" - and the variables reflecting social status "where live" and "social class" are no longer significant and drop out of the model. Although, the incidence and prevalence of smoking has been associated with social deprivation and social class;, there is no supportive evidence that such a link may be through social class. While some increases in the parameter estimates are to be expected when adding new significant variables to logistic regression model, the large decrease in the parameter estimates confirms that the effect of the variables "age" and "which parent live with" have substantially reduced. This is consistent with socio-environmental variables having an intervening effect between age, parent(s) and smoking. Similarly, from model 2 to model 3 when socio-psychological variables are added to model 2 (see Table 2), a modest decrease in the parameter estimates of demographic and socio-environmental variables can be noted. This decrease is consistent with the socio-psychological variables having an intervening effect between demographic and socio-environmental variables and smoking. Out of the 15 socio-psychological variables all of which were significant in the bivariate analysis (Table 1) only 6 remain statistically significant and are included in model 3. However, the modelling process shown in Table 2 provides some indications that these 6 variables may owe their significance to other variables excluded from the analysis. It is interesting to note that self-esteem, worrying about school and health variables which are reported in the literature (see introduction) as correlates of smoking are not included in the final model with this data set.

Table 2 approximately here

4-Discussion of the results

The results from model 3 is shown in Table 3 which can be thought of as a standard multivariate analysis. It can be seen that, from these results, it is quite easy to conclude that "best friend" has the largest effect on smoking followed by the variables "have partner"; "how feel with opposite sex"; "which parent live with"; and the two "worrying" variables. The results in Table 3 also suggest that those who claimed they were happy with their body shape have a reduced risk of smoking to 80% of those who were not happy with their body shape. Furthermore, those who claimed to be at ease with the opposite sex appear to have an increased risk of smoking by a factor of nearly 2 compared to those who claimed to be very uneasy. Similarly those whose best friend was a smoker have an increased risk of smoking by a factor of 14 compared to those whose best friend was not a smoker.

Table 3 approximately here

Moreover, those who live with foster parents appear to have an increased risk of nearly 4 times that of pupils who live with both parents. This latter result can be explained as a selection bias; it is plausible that smoking may well have started while in care prior to placement with foster parents. However, the model fitting results of Table 2 suggest that whilst the foster parents effect can be explained, the results for the remaining variables may be more complicated to interpret and certainly may not be interpreted as a direct effect on smoking.

Table 2 suggests that subjective variables may be acting as intervening rather than independent variables. It is interesting to note that the variable "whether happy with weight", significant on its own (Table 1) failed to remain in the final model after controlling for other variables. On the other hand, the variable "being happy with body shape" remains significant and enters the final model. It is therefore plausible that the variable "being happy with body shape" may reflect the respondents state of mind at the time of the interview regardless of the actual body shape and weight. This means that, in addition to body shape, the variable "happy with body shape" may well be an indication of this group's insensitivity to social pressures; social influences that may lead to smoking (see Background). In particular individuals with similar body shape may have different outcomes; some become smokers and some would not.

The results for the variable "best friend smokes" in Table 2 suggest that this variable may be an intervening variable. It can be seen that the change in its parameter estimate from model 2 to model 3 is over two times its standard error. If there is a 'true' best friend effect it is too complex to distinguish with cross-sectional data. As discussed in the introduction, it is plausible that prior to taking up smoking such pupils may have had a lower self-esteem, a wish to gain confidence, a desire to belong to a peer group and possibly lacked social and parental guidance. There is some evidence to suggest that parental influence indirectly predicts lower levels of smoking;;. Therefore, the "best friend" effect may not be straight forward to interpret with these data because we have no knowledge of these pupil's previous smoking habits; they may have been a smoker prior to the friendship. Furthermore, the high odds ratio for 'best friend' may well be inflated due to a pragmatic sampling strategy. Surveying of all pupils within classrooms in the selected schools may have achieved maximum response rate and overcome the practical issues such as increasing sample size due to schools dropping out, but, has added complexities to data e.g (i) the clustering effect where pupils with similar outcome characteristics tend to form clusters (or social groups), and (ii) the possibility of double counting where pupils in the same classroom citing each other as best friend (e.g see Stuart). Cross-sectional analysis of such data will lead to artificially inflated effect of "'best friend" on smoking (e.g see Davies).

The result for the variable "how feels with opposite sex" in Table 3 suggests that feeling at ease with the opposite sex increases the likelihood of being a smoker which may imply that "the more confident" pupils are likely to smoke. On the other hand, Table 2 suggests the possibility of a more dynamic social process (notice changes in the parameter estimates' of model 3 compared with model 2). Once again we have no knowledge of neither the pupils past smoking behaviour nor past patterns of their social interactions with the opposite sex. Therefore it would be unwise to interpret the result for this variable simply as is, and certainly it should not be confused with any causal relationship between confidence and smoking.

From the results in Table 2 and 3 similar conclusions can be made for the variables "considers health when choosing food", "worry about money problems" and "worry about family problems". These three variables were statistically significant and remained in the final model. As most pupils have been made aware of the dangers of smoking, these variables may be a proxy for the underlying effect of attitudes to health and smoking. The choice of food represents the health consciousness of pupils suggesting that those who attach importance to health related behaviour have a reduced risk of being a smoker. While the worrying variables serve to demonstrate the subjective effect of smoking where smoking leads to the maintenance of smoking;. Again without prior information on the pupils smoking behaviour these results do not constitute evidence for worrying leading to smoking. This association between smoking and worrying/health may help explain the prevalence but not incidence of smoking.

Finally, it appears that those pupils who have or had a partner (boy friend or girl friend) are more likely to be a smoker than those who never had a partner. It is not clear whether having a partner is the reason for pupils' smoking habits, or, whether young people smoke to be sociable. However, from the results shown in Table 2, while there is a small change in the parameter estimate of the objective variables from model 2 to 3, the change in the parameter estimates of the socio-environmental variables are around twice their standard error. It is, therefore plausible that the variable "have a partner" may be reflecting a "best friend" effect or a "foster parent" effect which were discussed earlier.

Concluding comments

With survey or observational studies it is possible to measure a wide range and mix of individual, social and environmental variables, however, it is not always possible to measure some other variables such as personal attributes including biological and psychological variables. Thus leading to the complex issues arising from omitted variables and subjectivity/objectivity. Cross-sectional multivariate analyses of such data will almost certainly lead to a mis-specification error and hence to erroneous results. Caution must be exercised when interpreting the results from survey or observational studies even when they are based on multivariate analyses. These issues may be addressed with the application of instrumental variables within a longitudinal study framework. Indeed, longitudinal study designs are essential to address some of these issues; including studies of patterns of incidence (e.g. see Gilpin, why do young people start smoking; and prevalence of smoking in relation to socio-psychological well being, communication problems between young people and their informal and formal support network (e.g. see Flay), and dynamics of smoking, for example, a high risk activity may or may not follow cessation of smoking.

Acknowledgement

The data for this paper was provided by the kind permission of Public Health Directorate, Northern & Yorkshire NHS Executive, John Snow House, Durham University Science Park, Durham DH1 3YG.

Tony Goodall was the coordinator of Yorkshire Heartbeat (Public Health Department) who initiated the health behaviour related survey in Yorkshire, which was carried out by the Schools Health Education Unit at Exeter University.

References

1. Shahtahmasebi S. Statistical modelling of dependency in old age [Ph.D]: University of Wales, Bangor, North Wales; 1995.

2. Shahtahmasebi S, Davies R, Wenger C. A longitudinal analysis of factors related to survival in old age. The Gerontologist 1992;333:404-413.

3. Wenger GC, Davies R, Shahtahmasebi S. Morale in Old Age: Refining the Model. International Journal of Geriatric Psychiatry 1995;10:933-943.

4. Davies RB. From cross-sectional to longitudinal analysis. In: Dale A, Davies RB, editors. Analysing social and political change: A casebook of methods. London: Sage; 1994. p. 20-40.

5. Pletcher JR, Schwarz DF. Current concepts in adolescent smoking. Curr Opin Pediatr 2000;12(5):444-9.

6. Tyas SL, Pederson LL. Psychosocial factors related to adolescent smoking: a critical review of the literature. Tob Control 1998;7(4):409-20.

7. Jensen EJ, Overgaard E. Investigation of Smoking Habits among 14-17-year-old Boarding School Pupils: Factors which influence Smoking Status. Public Health 1993;107:117-123.

8. Hine C. Report on stopping or reducing smoking: issues for Health Authorities: Health Care Evaluation Unit. Department of Epidemiology and Public Health Medicine. University of Bristol; 1992.

9. Lader D, Matheson J. Smoking among secondary school children in 1990. An enquiry carried out by Social Survey Division of OPCS on behalf of the Department of Health, the Welsh Office and the Scottish Office Home and Health Department: OPCS, Social Survey Division. London:HMSO; 1991.

10. Goddard E. Why children start smoking. An enquiry carried out by the Social Survey Division of OPCS on behalf of the Department of Health: OPCS, Social Survey Division London:HMSO; 1990.

11. Bolling K. Smoking among secondary school children in England in 1993. An enquiry carried out by Social Survey Division of OPCS on behalf of the Department of Health: OPCS. Social Survey Division London:HMSO; 1994.

12. Sussman SC, Dent W. Group self-identification and adolescent cigarette smoking: a 1-year prospective study. Journal of Abnormal Psychology 1994;103(3):576-80.

13. Murphy NT, Price CJ. The influence of self-esteem, parental smoking and living in tobacco production region on adolescent smoking behaviour. Journal of School Health 1988;58(10):401-5.

14. Parsons JT, Siegel AW, Cousins JH. Late adolescent risk-taking: effect of perceived benefits and perceived risks on behavioural intentions and behavioural change. Journal of Adolescence 1997;20:381-392.

15. Moore S, Gullone E. Predicting adolescent risk behaviour using a personalized cost-benefit analysis. Journal of Youth and Adolescence 1996;25(3):343-359.

16. Smith AMA, Rosenthal DA. Adolescents' perceptions of their risk environment. Journal of Adolescence 1995;18:229-245.

17. Upadhyaya HP, Deas D, Brady KT, Kruesi M. Cigarette smoking and psychiatric comorbidity in children and adolescents. J Am Acad Child Adolesc Psychiatry 2002;41(11):1294-305.

18. Shek DT. Family environment and adolescent psychological well-being, school adjustment, and problem behaviour: a pioneer study in a Chinese context. Journal of Genetic Psychology 1997;158(1):113-28.

19. McNeill AD, Jarvis M, West R. Subjective effects of cigarette smoking in adolescents. Psychopharmacology-(Berl) 1987;92(1):115-7.

20. Clayton S. Gender differences in psychosocial determinants of adolescent smoking. Journal of School Health 1991;61(3):115-20.

21. Paavola M, Vartianian E, Puska P. Predicting adult smoking: the influence of smoking during adolescence and smoking among friends and family. Health Education Research 1996;11(3):309-315.

22. McNeill AD. The development of dependence on smoking in children. Psychopharmacology-(Berl) 1991;92(1):115-7.

23. Stead M, Hastings G, Tudor-Smith C. Preventive adolescent smoking: a review of options. Health Education Journal 1996;55:31-54.

24. Moolchan ET, Ernst M, Henningfield JE. A review of tobacco smoking in adolescents: treatment implications. J Am Acad Child Adolesc Psychiatry 2000;39(6):682-93.

25. Giovannucci E. An updated review of the epidemiological evidence that cigarette smoking increases risk of colorectal cancer. Cancer Epidemiol Biomarkers Prev 2001;10(7):725-31.

26. Balding J. Young People in 1991: Schools Health Education Unit, University of Exeter; 1992.

27. GLIM. Generalised Linear Interactive Modelling. In. 4 ed. Oxford, U.K: NAG; 1993.

28. Conrad KM, Flay BR, Hill D. Why children start smoking cigarettes: predictors of onset. British Journal of Addiction 1992;87:1711-24.

29. Glendinning A, Shucksmith J, Hendry L. Social class and adolescent smoking behaviour. Social Science and Medicine 1994;38(10):1149-60.

30. Oakley A, Brannen J, Dodd K. Young people, gender and smoking in the United Kingdom. Health Promotion International 1992;7(2):75-88.

31. Royal College of Physicians. Smoking and the Young. London; 1992.

32. Charlton A. The Brigantia Smoking Survey: a general review. Public Education About Cancer. UICC Technical Report Series 1984;77:92-102.

33. U.S. Department of Health and Human Services. Preventing Tobacco Use Among Young People. A Report of the Surgeon General: Public Health Service, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health; 1994.

34. Eiser J, Morgan M, Gammage P, Gray E. Adolescent smoking: attitudes, norms and parental influences. British Journal of Soc. Physio. 1989;28:193-202.

35. Stuart A. The idea of sampling. High Wycombe: Charles Griffin and Company Ltd; 1984.

36. Davies RB. The limitation of cross-sectional analysis. In: Crouchley R, editor. Longitudinal Data Analysis. London: Sage; 1987.

37. Gilpin EA, Pierce JP, Rosbrook B. Are adolescents receptive to current sales promotion practices of the tobacco industry? Preventive Medicine 1997;26(1):14-21.

38. Flay BR, Hu FB, Siddiqui O, Day LE, Hedeker D, Petraitis J, et al. Differential influence of parental smoking and friends' smoking on adolescent initiation and escalation of smoking. Journal of Health & Social Behavior 1994;35(3):248-65.

Appendix I

Definitions:

Smokers: are those who claimed to smoke either regularly or occasionally; non-smokers are those who either never smoked or had given up smoking.

Missing Values Category: this category was created if the proportion of non-response (missing values) for any question was greater than 1%, otherwise, missing values re-category was assigned to the category with the highest cell frequency.

The following variables are created from the data set );

Social class - the only measure of class available from the survey is crudely based on the type of paper read by the household of the respondent: High = Quality Broadsheet newspapers (e.g. Times, Guardian, Telegraph, Independent, Scotsman, Observer); Medium = Quality Tabloid news paper (e.g. Mail, Express, Today, Scottish Daily Express); Low = Popular Tabloid News Paper (e.g. Star, Mirror, Sun, Scottish Daily Record, Sunday Sport, Sunday People).

Self-esteem - this variable is based on 9 statements each allowing a score of 0-2; "0" agree, "1" not sure and "2" disagree. The total will take a value of 0-18, of which a score of 0-4 is low, 5-9, and 10-14 are "lower middle" and "upper middle" respectively, and 15-18 is high self-esteem. The statements are:

"There is lots of things about myself that I would like to change."

"When I have something to say in front of teachers, I usually feel uneasy."

"I often fall out with other pupils at school."

"I often feel lonely at school."

"I think other pupils usually say nasty things about me."

"When I want to tell a teacher something I usually feel shy."

"I often have to find new friends because my old ones are with somebody else."

"I usually feel foolish when I have to talk to my parents."

"I feel comfortable talking to other pupils at school."

Appendix II

For this project we are interested in the prevalence of smoking s as opposed "others", for this purpose the smokers are defined and coded "1" as those who classed themselves as occasional and regular smokers, and the non-smokers as those who never smoked, had tried it once or twice, and those who had given up smoking and are coded "0".

To fit a normal regression to this sort of data where the response is a binary variable (1/0 indicating yes/no, or, success/failure) data needs to be tranformed. Generally a suitable transformation is the logit (see reference 9) often referred to as logistic regression with the probability density function given by:

One of the advantages of the logit model is that we can estimate the probability of "smoker" or "high alcohol consumption" using the prevalence rate in the sample. Alternatively, the odds of being a smoker can be derived as follows:

let x1i represent sex; x1i=0=male, x1i=1=female and let β1 be the logistic coefficient of x1i:

logit (p1; y=smoker|female) = α + β1 (x1i=1) + β2x2i + ...

i.e the log odds of being a smoker given a female respondent, while the log odds for a male respondent is given by

logit (p2; y=smoker|male) = α + β1 (x1i=0) + β2x2i + ...

because β1 is the difference between the two logs, i.e.

logit (p1) - logit (p2) = β1 Þ log (p1/p2) = β1

Þ log (Odds Ratio) = β1 Þ OR = exp(β1 )

The confidence limits (CLs) for the OR can easily be obtained from the standard errors of the logistic coefficients. The odds ratios with their appropriate confidence limits are given in Tables 4.

 

 

 

 

 

 

Table 1 - Selected variables from the Health Related Behaviour Questionnaire thought to be associated with smoking habits of young people. Bi-variate cross-classification; N=9230, * p<.05, **p<0.001, ***p<0.00001.

Explanatory variables

Smoking

Socio-demographic

p

1- Age

***

2- Gender

***

3- Which parents live with

***

4- Where live

***

5- Social class (see 3)

*

Socio-environment

6- Whether smoker

-

7- Whether drinks

***

8- Whether at least one in family smokes

***

9- Whether relative smokes

***

10- Whether best friend smokes

***

Socio-psychological

 

11- Self-esteem

**

12- Whether happy with weight

***

13- Whether happy with body shape

***

14- Whether considers health when choosing food

 

14- Whether have a steady partner

***

15- How feels when meeting opposite sex

***

16- How feels when meeting own sex

***

17- Have a drink when have problem

***

How much worry about:

 

18- school problems

***

19- money problems

***

20- health problems

***

21- career problems

***

22- problems with friend

***

23- family problem

***

24- the way you look

***

Table 2- Model fitting results for the model of smoking - N=9230

Model 1

Model 2

Model 3

Explanatory variables

Parameter

estimate

Standard error

Parameter

estimate

Standard error

Parameter

estimate

Standard error

Demographic factors

 

Age

12-13

14-15

0.00

1.06

 

0.06

0.00

0.68

 

0.07

0.00

0.61

 

0.08

Gender

male

female

0.00

0.52

 

0.06

0.00

0.68

 

0.07

0.00

0.67

 

0.08

Which parent live with

both parents

mother only

father only

mother & step-father

father & step-mother

foster parents

other

0.00

0.47

0.69

0.80

0.77

1.63

0.73

 

0.08

0.18

0.09

0.20

0.29

0.19

0.00

0.23

0.49

0.53

0.44

1.56

0.38

 

0.10

0.23

0.11

0.25

0.37

0.24

0.00

0.12

0.37

0.39

0.27

1.32

0.28

 

0.11

0.24

0.11

0.26

0.38

0.25

Where live

town/city centre

town/city suburb

small town/city centre

small town/city sub.

in village

outside town/village

0.00

-0.30

-0.08

-0.26

0.10

0.20

 

0.13

0.11

0.11

0.10

0.13

 

     

Social class(see Appendix)

high

medium

low

0.00

0.15

0.25

 

0.10

0.09

       

Socio-environmental factors

 

Whether drinks

no

yes

 

 

 

0.00

1.11

 

0.08

0.00

0.94

 

0.08

Whether at least one family smokes

no

yes

   

0.00

0.60

 

0.08

0.00

0.47

 

0.08

Best friend smokes

no

yes

   

0.00

2.81

 

0.07

0.00

2.65

 

0.07

Socio-psychological factors

 

Have partner

never had one

not at the moment

yes, few weeks

yes, up to 6 months

yes, up to a year

yes, > 1 year

       

0.00

0.70

1.18

1.56

1.03

1.15

 

0.17

0.19

0.20

0.25

0.21

How feel with opposite sex

very uneasy

a little uneasy

at ease

       

0.00

0.28

0.64

 

0.14

0.14

Happy with body shape

no

yes

       

0.00

-0.22

 

0.08

Considers health when choosing food

never

sometimes

quite often

very often

always

       

0.00

-0.48

-0.73

-1.15

-0.83

 

0.11

0.12

0.15

0.19

Worry about money problems

never/hardly ever

a little

quite a lot/a lot

       

0.00

0.29

0.54

 

0.09

0.09

Worry about family problems

never/hardly ever

a little

quite a lot/a lot

       

0.00

0.16

0.31

 

0.10

0.09

 

Table 3- Odds Ratios for the model of smoking prevalence with their appropriate 95% Confidence Limits after controlling for other factors

Explanatory variables

Lower

Odds Ratio

Upper

Age

12-13

14-15

1.00

1.59

1.00

1.84

1.00

2.13

Sex

male

female

1.00

1.68

1.00

1.96

1.00

2.30

Which parent live with

both parents

mother only

father only

mother & step-father

father & step-mother

foster parents

other

1.00

0.92

0.92

1.18

0.78

1.78

0.79

1.00

1.13

1.45

1.48

1.31

3.74

1.31

1.00

1.40

2.31

1.84

2.19

7.87

2.14

Whether drinks

no

yes

1.00

2.19

1.00

2.56

1.00

3.00

Whether at least one family smokes

no

yes

1.00

1.40

1.00

1.62

1.00

1.89

Best friend smokes

no

yes

1.00

12.23

1.00

14.16

1.00

16.36

Have partner

never had one

not at the moment

yes, few weeks

yes, up to 6 months

yes, up to a year

yes, > 1 year

1.00

1.44

2.25

3.21

1.72

2.10

1.00

2.02

3.25

4.77

2.80

3.17

1.00

2.85

4.71

7.09

4.53

4.78

How feel with opposite sex

very uneasy

a little uneasy

at ease

1.00

1.01

1.43

1.00

1.32

1.89

1.00

1.74

2.49

Happy with body shape

no

yes

1.00

0.69

1.00

0.80

1.00

0.93

Considers health when choosing food

never

sometimes

quite often

very often

always

1.00

0.50

0.38

0.23

0.30

1.00

0.62

0.48

0.32

0.44

1.00

0.77

0.62

0.43

0.63

Worry about money problems

never/hardly ever

a little

quite a lot/a lot

1.00

1.12

1.43

1.00

1.33

1.71

1.00

1.58

2.05

Worry about family problems

never/hardly ever

a little

quite a lot/a lot

1.00

0.97

1.15

1.00

1.17

1.37

1.00

1.42

1.62

 

 





This Article
Right arrow Abstract Freely available
Services
Right arrow Similar articles in this netprints
Right arrow Download to citation manager
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Shahtahmasebi, S.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Shahtahmasebi, S.
Related Collections
Right arrow Paediatrics:
Other Paediatrics

Right arrow Public Health:
Smoking


HOME HELP FEEDBACK BROWSE ARTICLES BROWSE BY AUTHOR