The concept of statistical significance cannot be underestimated because it is the means with which researchers can either nullify or validate their hypotheses. A statistically significant study is one in which the authors affirm that they possess approximately 95% confidence that the findings in their research were not a fluke.
Statistical procedures mentioned in the report
In essence the study under analysis was an epistemological research in which the authors were trying to ascertain the dangers of environmental exposure to a carcinogenic substance which in this case was second hand smoke. They carried this out through the selection of case group members from various parts of Europe; they were 650 participants. The latter members had been exposed to second hand smoke through three major channels which were the workplace, spouses and in their childhood and were lung cancer patients. They studied the prevalence of lung cancer in the case group and compared it to a control group that comprised of 1542 participants that had not been exposed to secondhand smoke. (International Agency for Research on Cancer, 1998)
In terms of statistics, a range of procedures were carried out, however, the most important was the calculation of the odds ratio (OR) in all the concerned groups i.e. for those who were exposed for over 15 years in childhood, those who were exposed to secondary smoke at their workplaces and among those exposed to it by their spouses. It should be noted that the odds ratio was the prevalence of lung cancer amongst the experimental group under analysis(y) versus prevalence of the control or unexposed group (z) as denoted by y/z. These ratios were calculated through two major pathways which included unconditional and conditional logistic regression for the exposure to spousal, childhood and workplace groups. The unconditional analysis was adjusted for center differences, age and sex differences. On the other hand, the conditional regression analysis was adjusted for center specific matching. All the odds ratios calculated entailed an analysis of the confidence intervals which would range from 0.64-1.7 and also analyses of P values for the selected parameters was done.
Conclusions reached in the study and whether the conclusions are appropriate
The conclusions reached in the study were that there was no linkage between exposure of children to second hand smoke and prevalence of lung cancer- a conclusion based on odds ratio findings of 0.78 for confidence intervals of 0.93-1.44. Furthermore, the researchers stated that there was minimal risk of lung cancer due to exposure to workplace environmental tobacco smoke (ETS). This was supported by the 1.17 odds ratio obtained at a confidence interval of between 0.94 and 1.45. Lastly, the researchers also concluded that there was minimal risk due to exposure to environmental tobacco smoke by spouses owing to odds ratio findings of 1.16 at confidence intervals of 0.93 to 1.44.
All the stated confidence interval were 95% percent so it can be said that the conclusions reached could be reasonably accurate given only a five percent margin of error. Furthermore, as stated earlier, an odds ratio of less than one indicates that prevalence of the matter under analysis ( lung cancer) is lower in the experimental population than the control population. Consequently, asserting that there was no correlation between second hand smoke exposure among children is accurate. In terms of exposure to tobacco smoke from spouses and also at the workplace it was found that odds ratio of 1.16 and 1.17 existed respectively. The latter percentages were quite small and it was therefore okay to assume that there was no significance between those parameters i.e. environmental tobacco smoke at the workplace/ from spouses and increased risk of lung cancer. In other words, their conclusions were appropriate.
Whether the findings are statistically significant, process used to make this determination and the level of significance
The findings were not statistically significant. In order to ensure that a study reaches statistical significance, it is often necessary to consider a number of things and the first one is the sample size. This research employed a total of 2,192 participants. An n value of greater than five hundred is often considered large enough and can therefore be fully relied upon. In other words, the confidence interval of 95% has been reached owing to the use of such a large sample size. Secondly, statistical significance is often denoted by the p value. Usually, when it is found that this value falls just slightly below 0.05 then it can be considered relatively significant.
However, if it falls way below the 0.05 mark then that would be a highly significant result. The calculations made by the authors in this research found that the p values were all above 0.05 so they could not be considered significant. (Chance and Rossman, 2005) For instance when assessing the risk due to workplace exposure of the case subjects, it was found that the odds ratio of 1.17 has a p value of 0.23 which implied no significance. Furthermore, when looking at the combined risk of lung cancer due to both workplace and spousal exposure at 1.14 odds ratio, it was found that a p value of 0.82 existed; not significant. It is often assumed in any research that the greater the p value, the lower the reliability that can be placed on a study. In this regard, since the researchers found very high p values, then the relation between the variables cannot be true. This means that the results are not significant and that there is minimal link between exposure to environmental tobacco and prevalence of cancer.
Although the latter study has generated a lot of controversy in the field, one cannot apportion blame to the authors of this study because they followed rigorous statistical analyses to reach their conclusion. A follow through of p values and confidence intervals in their paper gives support to a null hypothesis which states that there is no significant correlation between exposure to environmental tobacco and lung cancer.
International Agency for Research on cancer (1998). Multicenter case control study of exposure to environmental tobacco smoke and lung cancer in Europe. National cancer institute 90(19), 1441-1453.
Chance, B. & Rossman, A. (2005). Investigating statistical applications, concepts and methods. NY: Duxbury press.