A bayesian binary regression cohort model for analyzing the determinants of family planning in Bangladesh : With special focus on unmet contraceptive need A Bayesian Binary Regression Cohort Model for Analyzing the Determinants of Family Planning in Bangladesh with Special Focus on Unmet Contraceptive Need
A bayesian binary regression cohort model for analyzing the determinants of family planning in Bangladesh : With special focus on unmet contraceptive need
A Bayesian Binary Regression Cohort Model for Analyzing the Determinants of Family Planning in Bangladesh with Special Focus on Unmet Contraceptive Need
Huq Syeda Shahanara
ハック サイエダ シャハナラ
On Method<br /><br />The main focus of this thesis is to perform the cohort analysis of unmet contraceptive need data of several surveys in Bangladesh. Most of traditional studies on unmet contraceptive need have been done by employing classical methods, such as the logistic regression analysis. Our objective of the analysis is not only to see the effects of age, period and cohort separately but also to observe the interaction effects of age, period and cohort components with important covariate, which was impossible with any existing cohort analysis methods.<br /><br />We proposed a new Bayesian binary regression cohort model. Ours is the first model which incorporated the effects of explanatory variables (covariates) on the response variable. It is also true that our model is the first cohort analysis model which is capable of dealing with the survey data in the form of binary response variable. Our model not only helps examine the effects of age, period and cohort but also to explore the influence of the covariates on the response variable as well. Our model is applicable for unequally spaced survey data, because of the use of B-spline functions. The fit of models are evaluated by Akaike Bayesian Information Criterion (ABIC). The model which has the minimum ABIC value among the models fitted to the data is chosen as the best model.<br /><br /> Traditional statistical methods help examine the instantaneous influence of the covariates, but our method helps see the age-by-covariate, period-by-covariate and cohort-by-covariate effects on the response along with the instantaneous influence of the covariates. The performance of our Bayesian method have been checked by two simulation studies. The first set of simulation data is generated assuming no covariate effect, while in the second simulation, the covariate effects on the response variable is also envisaged as well as age, period and cohort effects. It is shown that our method can detect the true model in both cases. The reliability of the estimated parameters have also been inspected visually by attached confidence interval. A new way of inspection is devised to take into consideration the identification problem.<br /><br />Utilizing Information Criterion ABIC we can check whether two sets of data can be explained by the same model or not. It is realized by comparing the sum of ABIC values of the best models of separate analyses and ABIC value of the best model for combined data. An example is provided in Section 5.3.3, where it is concluded that mothers in rural areas are behaving differently from the mothers in urban areas.<br /><br />The present study considered only 'bi-variate' effects of covariates. But it is not difficult to modify our method to take into consideration the strength of covariates. This quantitative aspect of covariates will be included into our model in future.<br /><br />On Facts<br /><br />Combined surveys data on unmet contraceptive need and current contraceptive use are taken from a series of surveys such as CPSs 1983, 1985, 1989, 1991, and DHSs 1993-94, 1996-97 of Bangladesh (according to survey reports title). These surveys were not conducted with equally spaced intervals of time. But the application of our Bayesian binary regression cohort model to this type of data is straightforward. We apply our method to these survey data by utilizing B-spline function.<br /><br />Our model revealed interesting facts about contraceptive behavior of Bangladeshi women. Two response variables, UCN and contraceptive use, are analyzed by our proposed methodology. We revealed that:<br /><br />1. Women in the younger cohorts are more conscious about the total number of living <br /> children compared to older cohorts.<br />2. The field workers visit is increasing its efficiency.<br />3. Unmet contraceptive need increases among the mothers who have at most two<br /> children indicates that their tendency to stop child bearing increases as time <br /> progresses but this increasing behavior is not that much prominant their practice of<br /> Contraception.<br />4. In rural areas religion effect on UCN is lower in younger cohorts. Therefore, it can be <br /> concluded that younger generation of Muslim women in rural areas are changing their<br /> attitude towards "birth control" by adopting contraception. (This observation is <br /> obtained from the separate analysis of rural and urban data.) There are no cross<br /> effects between covariates and age, period or cohort effects on UCN of urban women.<br />5. Interaction effect of age by fieldworker's visit (i.e., the cross effect of women's age and<br /> the fieldworker's visit) on contraceptive use is clearly observed, but in very smaller<br /> magnitude.<br />6. It is an impressive findings that the use rate increases over time specially in rural<br /> areas.<br /> Evidences in favour of changings attitudes towards contraception among Muslims<br /> women are also observed.<br /><br />The findings of this study relating to contraceptive use are harmonious with our prior expectation and also with the published literature on this subject. Contraceptive use varies with age. Use rates are higher in the middle ages of child bearing. It can be said that women are likely to use contraceptive when they are in their middle ages. Lower use in the older ages may reflect the women's declining fecundity while lower use in the younger ages probably is due to want more child to reach their desired family size.<br /><br />Results of our analysis is encouraging in the sense that it is proving the effectiveness of female field force on the reduction of UCN. UCN could be reduced not necessarily interfering the personal intension about desirable size of the family. Of course, the reduction of the 'average' of desired family size is an important issue but it is out of the scope of the present study.<br /><br /> Family planning fieldworkers visit is an important covariate to reduce the unmet need not only by motivating women but also supplying the contraceptive methods to them. Some other studies have demonstrated that services of fieldworkers may be central to the success of the family planning program in a traditional society like Bangladesh. Bongaarts and Bruce (1995) mentioned that the best example of family planning performance under adverse condition is that of Bangladesh. In the invigorated national family planning program, an active cadre of female family planning health visitors reach into the community and home environments of women; the visitors make themselves available on a regular and predictable basis.<br /><br />Our method also pointed out that the women having unmet need are more concentrated in their middle ages. Fieldworker should encourage such women to accept the family planning method by explaining the advantage of small family. Also the percentage of women desiring additional children in future should be reduced, desired family size is one of the factor found to have an effect on contraceptive use in Bangladesh. Our model revealed that education reduces UCN in rural areas also. At this point it can be said that an improvement in the status of women should be helpful in reducing the UCN. Though "religious prohibition" is one of the important cause of UCN and which is normally not considered easily amenable to intervention. Our model provided that Muslim women in rural areas are gradually changing their attitude towards adoption of contraception to control family size over the period and this attitude is more stronger in younger cohorts.<br /><br />The family planning program should be made more explicit and purposeful in their efforts to reduce/close the gap between women's intention of child birth and contraceptive use. Program should extend it's activity specially in rural areas to reduce the areal variation between rural and urban areas of the country.