| In this study the theoretical basis, model fitting, variable selection, and hypothesis testing of Poisson Regression models and Bayesian approach were explored and applied to the cancer data in Ninawa province in Iraq to demonstrate the changes that have occurred of incidence cancer into three periods,1980-1990,1991-2000and2001-2010. In these periods Iraq has seen three big crises and wars, Iraq-Iran war (1980-1988), Gulf War (1990-1991), Economic sanctions (1990-2003) and American occupation (2003-2010). The data analyzed for each sub period using the Poisson regression with a log link function. The expected value (mean),//, of the number of incidence cases in Ninawa, which is a count variable and considered as response variable related to the factors:year, cancer site, age group, and sex which considered as the explanatory variables. The factor year group is a categorical predictor with11levels for the1980-1990,10levels for the period1991-2000, and10levels for the period2001-2010. Cancer site group is a categorical predictor with44levels represents all cancer sites according to the International Classification of Diseases (ICD9th). Age group is a categorical predictor with15levels represent five age groups from0-4to70+. Sex is a categorical predictor group with two levels.Due to detected overdispersion phenomenon, thus will give misleading inferences about the regression parameters of Poisson regression. An alternative method is Bayesian approach that was carried out to estimate the regression coefficients. Since no prior knowledge of the likely range of values of the parameters, therefore, the non-informative uniform prior distribution for regression coefficients are appropriate for this case, so we used the maximum likelihood estimates of the regression parameters as the starting values for the simulation. We used non-informative gamma prior distribution for the normal scale parameter, where the hyper-parameters (shape and the inverse scale) specified as0.001. The chain and seed are1. We generate2,000Gibbs pre-iterations to burn-in to ensure the convergence of parameters, and then generate10,000iterations for the posterior chain from the posterior distribution. Three graphical approaches and two statistical tests to assess convergence were conducted:The trace, autocorrelation, and density plots as well as Geweke test and effective sample size to ensure that the posterior distribution samples have achieved convergence before using them for Bayesian inference for the model parameters.We examined the time trend leukemia incidence rates by sex among age groups (0-19) and (20->70), during31years (1980-2010) to investigate the media reports of an increase in incidence of leukemia cancer in Ninawa province. In addition, we have investigated the media reports of an increase in incidence of leukemia cancer in overall Iraq among age group (0-14) years by sex.For Ninawa province, reduction in the cancer Incidence rate ratio in Ninawa in the third period (2001-2010) as well as in the second period (1991-2000). Our analyses found that breast cancer remained the first common cancer; while the lung, trachea, and bronchus the second in spite of decreasing as dramatically, while the highest increase was in leukemia in the third period (2001-2010) relates to the second period (1991-2000) but not to the first period (1980-1990). The cancer incidence rate ratio in men was decreased from more than33%than those of females in the first period (1980-1990), more than39%in the second period (1991-2000), and regressed to9.56%in the third period (2001-2010). Our results showed that there were increased in the number of cancer cases, but when it analyzed statistically with population growth in the Ninawa province, there were decreased in incidence rates in most cancer types.For the time trend leukemia cancer incidence in Ninawa province during the period from1980to2010, the analysis showed that the incidence rates varied markedly by age and sex. For the0-19age group, the incidence rates among females fluctuated without discernible pattern, the average number of leukemia incidence cases diagnosed each year during (1980-2010) increased from (0.4) to (8). Trend among males were more complex, fluctuated with no clear pattern, the average number of leukemia incidence cases diagnosed each year during (1980-2010) increased from (0.498) to (5.73). For the20->70years age group, among females, the average number of leukemia incidence cases diagnosed each year during (1980-2010) increased from (0.754) to (8.3). Among males, the average number of leukemia incidence cases diagnosed each year during (1980-2010) increased from (0.326) to (8.82). A study found that males aged0-19years are more sensitive than females from the first shock when an extremely major events occur, while females of the same age, sensitivity to a precarious situation or perilous position. For the20->70year’s age group, the consequences of economic crises and war influenced among males more than females. Probably the psychological factors, where the intense fear and the high risk which threatened the persons who more sensitive and vuhierable to the first shock was caused increase in the incidence rates, in addition to economic crises.For the time trend leukemia cancer incidence in overall Iraq, we found a statistically significant increasing trend in the incidence of childhood leukemia (0-14years of age), during the period1986-2010. The estimated annual percent increase1.51%among boys and2.02%among girls per year. The incidence rate was2.66per100,000for both sexes,3.08per100,000for boys and2.23per100,000. The incidence rates of childhood leukemia in Iraq similarity to the neighbors and developing countries and less than developed countries. |
PDF Full Text Request