Capítulo 3. Aislamiento y caracterización de una nueva isoforma de
3.3.5 Análisis de la estructura de los genes de H2A.Z en especies representativas
Report
Compared with the findings in the Year 2010 Survey report (a summary was pre- sented in Section 1.3.3), additional main findings of the drug-trying behaviour among young people in England by the study of the associations and relation- ships among drug-trying response variables and covariates (i.e. the smoking, drinking and drug-related socio-demographic variables) are summarized as fol- lows:
(1) Similar to the finding in the Year 2010 Survey report, results of the percentage contingency tables, box plots and polychoric correlation plots consistently show the strong positive association between smoking and drug-trying behaviour of the students in England and that there are different patterns of pairwise
CHAPTER 3. EXPLORATORY DATA ANALYSIS 88 associations between the smoking variables and the 15 individual drugs. Re- sults of the percentage contingency tables, box plots and polychoric correlation plots further reveal that the strong positive association between smoking and drug-trying behaviour of the students in England is highly contributed by the following smoking covariates: (1) the attitude of the students’ family towards smoking (CgFam1); (2) the students’ cigarette smoking status (CgStat1); (3) num- ber of cigarettes smoked by the students in the previous week (Cg7Num); (4) frequency of purchasing cigarettes from shops by the students (CgBuyF1); (5) sources of obtaining cigarettes by the students (CgGet); (6) whether there were smokers inside the students’ houses (CgWho1) as well as (7) the proportion of people a student knows who smoke (CgEstim).
(2) Similar to the smoking variable, results of the percentage contingency ta- bles, box plots and polychoric correlation plots are consistent to the finding in the Year 2010 Survey report that there is a positive association between drinking alcohol and drug-trying behaviour of the students in England and that there are different patterns of pairwise associations between the drinking variables and the 15 individual drugs. Results of the percentage contingency tables, box plots and polychoric correlation plots further reveal that the positive association be- tween drinking and drug-trying behaviour of the students in England is mainly contributed by the following drinking covariates: (1) the attitude of the stu- dents’ family toward drinking alcohol (AlPar1); (2) usual frequency of drinking alcohol by the students (AlFreq2); (3) sources of buying alcohol by the students (AlBuy); (4) whether there were drinkers inside the students’ houses (AlWho1); (5) types of incidences when the students drank alcohol (Al4W1) as well as (6) the proportion of people a student knows who drank alcohol (AlEstim).
(3) For the drug-related socio-demographic variables, results of the percent- age contingency tables, box plots and polychoric correlation plots support the
findings in the Year 2010 Survey report that the drug-related socio-demographic variables, namely (a) age of the students (Age); (b) how often the students had been excluded from schools (ExclAN1) and (c) how often the students played tru- ant (Truant1), are strongly and positively associated with drug-trying response variables. However, these three drug-related socio-demographic variables exert different patterns of pairwise associations with the 15 individual drugs. These three drug-related socio-demographic variables are particularly strongly corre- lated with the five drugs: (1) cannabis; (2) poppers; (3) cocaine; (4) ecstasy and (5) magic mushrooms.
(4) For the drug-trying response variables, results of the polychoric correlation plots show that the 15 drug-trying response variables are strongly and positively correlated with each other.
(5) The Year 2010 Year Survey report stated that "girls were less likely than boys to have taken drugs in the last year" (Fuller et al., 2011). According to the percentage contingency table in respect the gender variable (Gender), it reveals that the aforesaid statement is valid for seven drugs (cannabis, magic mush- rooms, crack, LSD, ketamine, anabolic steroids and tranquillisers) of which the proportion percentages of male students trying them were slightly higher than female students. On the other hand, for the other eight drugs (heroin, cocaine, methadone, ecstasy, amphetamines, poppers, gas and other drugs), results of the percentage tabulate show the opposite. Similarly, the Year 2010 Survey re- port stated that the school-level variable (percentage of pupils eligible for the free school meals) was not significantly associated with drug use in the survey. However, the percentage contingency table in respect of whether the students have enrolled in free school meal scheme (FSM1) indicates that the students involved in the free school meal scheme are more likely to try cannabis, heroin, cocaine, magic mushrooms, methadone, ketamine, gas and tranquillisers.
CHAPTER 3. EXPLORATORY DATA ANALYSIS 90
The above additional main findings in the exploratory data analysis of the working data set provide hints to justify our planned effort in this research as elaborated in Section 1.4.2. To enrich the understanding of drug-trying behaviour among young people in England, development and application of advanced statistical methodologies are needed to further investigate the inter- actions among drug-trying response variables as well as to further study the associations among drug-trying response variables and the smoking, drinking and drug-related socio-demographic variables in the working data set.
3.4
Summary
This chapter has summarised the results of the exploratory data analysis in re- spect of the working data set of this research. There were 25.48% of the students who had ever smoked, and 44.83% of the students who had ever drunk alco- hol. Most family members were either against or neutral towards smoking and drinking behaviour of the students. Most students knew surrounding people who either smoked or drank or took drugs, and most of them had lessons about smoking, drinking and drugs. Regarding the usage of drugs, cannabis was the most used drug, of which 9.06% of the students used it, whereas tranquillisers was the least used drug, of which only 1.85% of the students used it. A large number of the students had never tried drugs, but there were still a substantial number of the students who had tried drugs, including a few who had tried more than six drugs.
Regarding the pairwise associations between drug-trying response variables and covariates, except CgWhoSmo, gender and free school meal covariates, in general, most of the smoking, drinking and socio-demographic covariates were positively associated with drug-trying response variables. Drug-trying response
variables were also strongly and positively associated with each other.
Also, empty cells existed in some combinations of covariates and drug-trying response variables. This problem is needed to be addressed in Chapter 5 under logistic regression models.
When compared with the findings of the Year 2010 Survey report, examination of the pairwise associations and relationships in respect of drug-trying response variables and covariates (i.e. the smoking, drinking and socio-demographic vari- ables) of the working data set by percentage contingency tables, box plots and polychoric correlation plots shed additional light to help understanding more about the drug-trying behaviour of the students. These additional findings (as summarised in Section 3.3.5) were not found in the Year 2010 Survey report. In Chapter 4, we continue our analysis by investigating the missingness of the working data set. However, before such investigation, we discuss the missing data theory applied in the working data set in Section 4.1.
Chapter 4
Missing Data Theory, Methodology
and Application
4.1
Overview of Missingness
Missingness occurs for various reasons. For item non-response, reasons may include: (1) a respondent may not understand the question; (2) a respondent does not wish to answer the sensitive question; or (3) a respondent cannot figure out which option to choose in the case of multiple-choice questions. Moreover, if survey questions are deemed too tedious or too sensitive to answer, a respondent may refuse to answer (Tourangeau and Yan, 2007). Also, the internal routing system in a questionnaire may be another reason for item non-response. For unit non-response, a respondent may either have no interest or refuse to provide answers to the questionnaire or is unable to be interviewed due to language bar- rier and disabilities (Lavrakas, 2008).
Missing data are ubiquitous in societal and behavioural science studies (Lit- tle and Schenker, 1995), as well as in most medical, clinical and epidemiological research studies (White et al. (2009); Sterne et al. (2009); Tu and Greenwood (2012)), and are prevalent in large-scale surveys, including the "Health Survey
for England" and "Smoking, Drinking and Drug Use among Young People in England" survey series. De Leeuw et al. (2008) listed the possible causes of the increase in missing data in surveys, which include: (1) respondents are not having an answer to the question and (2) respondents’ refusal to provide a re- sponse. Regarding the Year 2010 Survey employed in this research, the probable causes of the missing data included the followings: (1) the survey questionnaire contained sensitive questions which the students refused to answer and (2) a portion of the students in the survey might possess insufficient information to answer some questionnaire questions (Kyureghian et al., 2011).
The problem of missing data is a major issue in statistical analyses. Schafer and Graham (2002) stated that since most statistical analyses are not designed to deal with missing values, the occurrence of missingness hampers the statis- tical analysis of scientific research. If missing data are not managed properly, missingness can lead to problems of bias in the statistical estimates and a loss of efficiency (White et al. (2009); Sterne et al. (2009); Carpenter and Kenward (2013)). Despite these problems, many researchers mistreat missingness by ei- ther treating missing values as merely another category or ignore the issue of the missing data and conduct a complete case analysis instead. In order to better understand the reason for the presence of missing data, there is a need to discuss the missing data theory and mechanisms, as well as methods to properly deal with missing data.