This thesis and the research work therein was carried out at the School of Mathematics, Statistics and Computer Science, University of Kwa Zulu-Natal, Westville Campus. Due to non-response in the survey, techniques used to handle missing data values were also investigated.
Introduction
Blood Donation
To ensure a safe blood supply in Africa, the Safe Blood for Africa Foundation (SBFA) was established as a multi-year, step-by-step implementation plan to establish the facilities and train the professionals needed to manage the millions of blood flows, tracking and testing. transfusions performed in sub-Saharan Africa (Volberding et al., 2008). According to Volberding et al. 2008), the SBFA Foundation programs are currently implemented in 18 African countries and are used to train more than 500 blood bank technicians annually, provide testing kits and supplies, and also provide technical assistance.
Mali
Individuals who are desperate for blood and for some reason cannot rely on replacement blood donors from their families often tend to pay for blood donations. The method each country uses to attract blood donors and meet its blood supply needs varies.
Data Collection and Descriptive Report
Data Collection
The questions asked by the designers are quite amazing to have a bird's eye view of KAP on blood donation and to plan for the improvement of blood supply. Techniques used to deal with missing data will be explored in later chapters and applied to this study.
Descriptive Report
When exploring blood donation and an appropriate way to give blood, Figure 2.5 shows that the majority of respondents reported that voluntary, unpaid (unpaid) donation is the appropriate way to give blood. When asked whether respondents have seen or heard messages about blood donation, it is striking that the majority of respondents on all three sites indicate that they have heard/seen messages about blood donation.
Logistic Regression
The Logistic Regression Model
Since estimation is made for the unknown probability for any given linear combination of the independent variables, a function is needed to connect the independent variables. Since it links the random and systematic components of the linear model, A is known as the link function (McCullagh & Nelder, 1989).
Odds Ratio
Hosmer and Lemeshow (2002) reported that a confidence interval estimate for the odds ratio can be obtained by calculating the endpoints of the confidence interval of the coefficient # and then calculating the exponents of these values. The basis for interpreting all logistic regression results is the relationship between the logistic regression coefficient and the odds ratio.
Parameter Estimation
The variances and covariances of the estimated coefficients can be obtained from the inverse of the observed information matrix, which is denoted as VarM. It follows that the variance-covariance matrix of | and #n in the univariate case is N, which is the inverse of the information matrix.
Goodness of Fit Test
- Pearson Chi-Square Statistic and Deviance
- The Hosmer-Lemeshow Test
- Area under the ROC Curve
- Information Criteria
- Measure of Association
Furthermore, for the deviation, it follows that Æ is the likelihood ratio test statistic of a saturated model with parameters versus the fitted model with. Agresti (2007) reports that the optimal model is the one whose adjusted values are closest to the actual outcome probabilities.
Overdispersion
If > is different from 1, then the distribution of the data is neither binomial nor Poisson, so an alternative distribution can be used. This lack of homogeneity can occur between individual groups or within individual observations (Olsson, 2002).
Logistic Regression Diagnostics
Let ℎ denote the diagonal element of the matrix ç defined in equation 3.6.1, then it follows. They argue that the change in the value of the estimated coefficients is analogous to Cook's proposed measure of linear regression.
Probit and Complementary log-log Models
- Probit Regression Model
- The Complementary Log-Log Model
The logit model uses the cumulative logistic function, while the probit model uses a standard cumulative functional form of the normal distribution. The probit model is generally estimated by maximum likelihood as is the case with the logit model.
Application
Similarly, a single unit increase for an individual who thought of spreading blood donation messages through organizations was 1.124 times more likely to be a donor compared to an otherwise identical individual (OR: 1.124, p-value = 0.7749, CI. According to in terms of blood donation practice, an individual who believed that blood donation is a good practice, important and that everyone should donate is approximately 1.25 times more likely to be a donor compared to the same individual who did not have this opinion.Nonsignificant the result means that, controlling for all other variables, the probability of being a one-unit donor increases for an individual who has a certain opinion about the appropriate way to donate blood, i.e.
Summary
One of them is gender and the other is knowledge about the different blood types. Regarding blood donation and knowledge about the different blood groups, a single unit increase for an individual who had knowledge about the different types of blood groups increases the chance of being a donor. All other explanatory variables did not seem to have a significant effect on the outcome of a donor in this region.
Correspondence Analysis
- Introduction to Correspondence Analysis
- Multiple Correspondence Analysis
- Adjustments to Inertias in MCA
- Application
- Summary
It can be seen that only 13.3 percent of the data is explained by the MCA map which is relatively low. More than 70 percent of the total variation is also accounted for by the first three dimensions. The variables located around the origin are not well represented in the map and do not contribute to the interpretation of the display.
Missing Data
Missing Data Mechanism
- Missing completely at random (MCAR)
- Missing at random (MAR)
- Not missing at random (NMAR)
The MAR mechanism depends on observed outcomes and possibly covariates, but not on unobserved outcomes, suggesting that the missing value may depend on observed data but not on unobserved data. If, conditional on the observed data, the probability of observing WC is independent of the WC values that would have been observed, but this probability is not necessarily independent of 0C and the observed WC values, then the missing WC values are MAR. The cause of missing values can essentially depend on the observed data as well as the unobserved data.
Ad Hoc Techniques to Deal with Missing Data
- Deletion Procedures
- Single Imputation Methods
One of the main differences between single imputation and multiple imputation is that while single imputation generates a single replacement value for each missing data point, multiple imputation creates multiple copies of the data set and imputes each copy with different plausible estimates of the missing value. This method can be biased as it overestimates the correlation between variables and underestimates the variability of the data. Maximum likelihood also plays a central role in missing data analyzes and is one of two approaches that methodologists currently consider state of the art (Schafer & Graham, 2002).
Maximum Likelihood (ML)
This means that the missing value is replaced with an observed value taken from a matched observation based on the non-missing variables. Enders (2010) reports that this imputation technique generally preserves the univariate distributions of the data and does not attenuate the variability of the imputed data to the same degree as other imputation techniques. As an example, SAS Global Forum (2012) considers ¼ observations with complete data and − ¼ observations with missing data at 0 and 0&.
Multiple Imputation
The likelihood can then be maximized to get ML estimates of: using the usual applicable techniques. 2010) report that unknown missing data replace ¼ independent simulated sets of values drawn from a posterior distribution of missing data conditional on the observed data. In summary, the idea behind the MI procedure is to use the distribution of the observed data to estimate a set of plausible values for the missing data. The advantage of the MI technique is that it can be applied to almost any type of data or model, and the analysis can be performed using any common software.
The Expectation-Maximization Algorithm
Subset Correspondence Analysis
Greenacre & Parbo (2006a), using a version of the same concept described in the aforementioned section, describe s-CA as an adaptation of CA. They claim that the aforementioned theory can be applied to a subset of the table, keeping the same row and column weights as in classical CA, but applied to a subset of profiles rather than a subset of the original table. Further assume that the corresponding subset of the mean vector is denoted by $, where $ is the weighted average of rows #∶ #&=$.
Application
It can be seen that only 257 cases were used in the analysis; in other words, approx. 20% of cases in the Mali dataset excluded from analysis due to missing data. The increase in precision is an indication of superior efficiency and statistical power achieved for the analysis of the FCS imputed data. It therefore follows that the two-dimensional number accounts for 79.6 percent of the variability in the data, and 20.4 percent is not included.
Summary
Results obtained from the FCS imputed model confirm those found in Chapter 3 which were based on logistic regression using the default CC method. The standard errors of the analysis of the FCS imputed data were smaller than those of the CC analysis leading to greater precision of the estimated coefficients. The increase in this accuracy is indicative of superior efficiency and statistical power obtained for the analysis of the FCS imputed data.
Conclusion
Malaria
WHO & Unicef (2003) have reported that malaria remains a major obstacle to health in sub-Saharan Africa, where it often takes its greatest toll on very young children and pregnant women, who are most at risk of malaria morbidity and mortality . Nelson & Williams (2007) have reported that malaria is prevalent in areas where child malnutrition is common and they argued that by improving child nutrition, malaria morbidity and mortality could be significantly reduced. Nelson & Williams (2007) have reported that severe anemia from malaria requires blood transfusion, but there is still an additional risk of the consequences of malaria due to the dangers of TTIs.
Syphilis
They further reported that although malaria can affect anyone, it is primarily a disease of the poor and the uninformed and the disease is much higher in the poor rural areas of Africa than in the developed urban areas. He further reported that in some chronic malaria patients, the parasite may not be seen or visible on the peripheral smear examination, but a unit of their blood will pass enough parasites to the recipient.
HIV/AIDS
Nelson & Williams (2007) reported that Mexico experienced a significant incidence of HIV infections among paid plasmapheresis donors in the 1980s when donors were infected by contaminated blood collection equipment during donation. On the other hand, Nelson & Williams (2007) reported that despite ensuring the safety of the blood supply in industrialized countries, blood transfusions still carry a significant risk of HIV transmission in many developing countries. Nelson & Williams (2007) further reported that new donors are much more common in developing countries than returning donors and that new donors and paid donors carry a much higher risk of transfusion-transmitted HIV, Hepatitis B Virus (HBV) and Hepatitis C. Virus infections ( HBC) in most populations.
Hepatitis B Virus (HBV)
Nelson & Williams (2007) report that HBV can be transmitted through percutaneous exposure to blood, sexual intercourse and from mother to infant. Nelson & Williams (2007) reported that blood donor screening for HBsAg was introduced in 1973. The introduction of HBsAg screening tests reduced the risk of HBV transmission by transfusion, as reported by Nelson & Williams (2007).
Hepatitis C Virus (HVC)
Main-Effect Model
Model Fitting
Plots in SAS
Class Age Gender RB KDBG HSMBD Edu_level Btrt BENote BMal SmsgsBD GP AppWay_VNRBD AppWay_PD/ param=ref;. Model Donor = Age Gender RB KDBG HSMBD Edu_level Btrt BEmerg BMal SmsgsBD GP AppWay_VNRBD AppWay_PD/ link=logit alpha=0.05;. Model Donor = Age Gender RB KDBG HSMBD Edu_level Btrt BEnote BMal SmsgsBD GP AppWay_VNRBD AppWay_PD/scale=none;.
Checking of Link Function
Multiple Correspondence Analysis
Missing Data
Don't know Q2.12 How did you experience the staff? who helped you with your last blood donation. Pregnant women 5 Vulnerable groups 6 Healthy people 7 Sick people 8 Don't know 9 Menstruating women 10. Being paid as a client 4 Don't know 5 Other (explain) 6. Question 5.2 What information must be provided to donors in advance of donating.
