Sample Research Proposal On Using Zero Inflated Poisson Models To Minimize Search Costs By Determining Extra Zero Counts

Abstract

Using Zero Inflated Poisson Models to Minimize Search Costs by Determining Extra Zero Counts
The Bay islands in Honduras are popular vacation places. The qualities that attract people to the Bay Islands in Honduras are the abundant water supplies. There are a number of water activities that can be conducted for re creational activities. One of the most popular activities is diving. The diving aficionados are drawn to the crystal clear water and the reefs. Other activities that tourist like to participate in the Bay Islands are the backpacking and hiking activities. The Bay Islands, Honduras is a tropical paradise that manifests the second most expansive underwater reef in the world. In addition to the recreational activities, the surroundings of the Bay islands and the North coast of Honduras are located in a tropical setting (Welcome, 2013).
In order for visitors to arrive at the wonderful tropical country of Honduras, there is no need for the tourists to receive mandatory vaccinations. The vaccinations are required due to the large mosquito populations that are found in the reservoirs around Honduras. These reservoirs are breeding grounds for mosquitos. The mosquitos are known transmitters of dengue and malaria. Malaria is an illness that is frequently encountered on the Northern coast of Honduras and in the Bay Island area. Malaria can be prevented by the ingestion of chloroquine. Chloroquine should not be ingested over long periods of time (Welcome, 2013).
Dengue is an illness that is encountered throughout Honduras. There are three distinct categories of dengue. These categories of dengue are: Dengue shock symptom, dengue hemorrhagic and dengue fever. In the event that the dengue symptoms are not properly diagnosed and treated, there is the potential of the dengue illness attaining distinct levels of intensity that vary from dengue fever to dengue shock syndrome. The standard treatment is to rest well, ingest substantial amounts of fluids and the ingestion of pain relievers that include acetaminophen (Welcome, 2013).
The objective of the research proposal is to examine the incidences of dengue, dengue hemorrhagic and dengue shock syndrome in Honduras in relation to the climate. The zero inflated Poisson model will be applied in order to minimize the search expenses that demonstrate the capacity of illustrating the covariates that are present in the manifestation of dengue. The zero inflation Poisson paradigms have the capacity of accommodating the additional zeros, autocorrelation and heterogeneity that is demonstrated in the dengue transmitted by the mosquitoes. The zero inflation Poisson methods are comprehensive paradigms that are applied to a variety of discrete chronological sequences that include the dengue data (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
This evaluation derives cost effective methods of comprehending the connection between the climate and the dengue correlated mortalities. The outcomes demonstrated that a deviance of the normal characteristic that is assessment of the extreme variation from the conventional fortnightly averages is substantially connected with the risk of contracting dengue from the mosquitoes. The periods that manifest an elevated temperature deviation are connected with the enhanced risk of contracting dengue in comparison with the periods where the temperature qualities have sustained a constant quality. These outcomes are in alignment with the hypothesis that hypothermic conditions are causal attributes to the enhanced risk of contracting dengue (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
The characteristics of the zero inflated Poisson (ZIP) methods supply a method of being able to model information that has a substantial amount of zeros. The variable that is designated as the response variable is formulated as a combination of the Poisson distribution and the Bernoulli distribution. In Lambert (1992), there was a presentation of an application that manifested defects in the processes of manufacturing in cases where it was inferred that the subtle non observed modifications in the surroundings were the causal attributes of the procedure migrating in a random manner between an imperfect condition where the manufacturing defects have the potential of occurring and are not inevitable to the perfect condition in which the manifestation of defects has a low incidence of occurring (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
The presence of the perfect condition enhanced the quantity of zeros in the information. In the zero inflated Poisson regression, a reaction vector of the measurements yi, considering that i= 1, 2, n, have an autonomous quality. The covariate matrices that are presented by yi = 0having a probability 1- p(x) in addition to y~ Poisson (λ) and having a probability that is demonstrated as p (x). The mathematical model is represented as:
1- p(x) + p(x) exp (-λ (z)) = P (yi = 0│x, z),
p(x) exp (-λ(z))λ(z)q x [q!(1- exp(-λ(z)))]-1, q= 1, 2.(Lambert, 1992).
This mathematical formulation that was demonstrated by Welsh (1996) enhances the potential of the probability of the zero result and causes the descaling of the residual probabilities in order that all of the probabilities have the characteristic of aggregating to one. The λ (z) and the p (x) functions are fashioned by log linear and logit functionalism respectively. The qualities of the covariates of x are not limited to being similar to the covariates of z. The formulation of the hurdle paradigm has a number of similarities with the zero inflation Poisson model. In both cases the zero inflation Poisson model and the hurdle model blends the Poisson distributions with the binomial probabilities. In the circumstance of the hurdle model, the zero category is maintained separately from the non – zero category by applying a truncated Poisson distribution to the non- zero yi. This is distinct from the zero inflation Poisson model in which the zeros take place in the perfect condition that has the probability of 1-p(x)) and in the imperfect condition that has the probability of p(x) exp (- λ (z))) (Welch et al., 1996).
An estimation model algorithm was developed by Lambert (1992) that optimizes the probability of the log – likelihood in the application of the zero inflation Poisson method. This process requires that adapting of two linear fitting paradigms (General linear model) a Poisson regression method and a logistic method. The mathematical relationship for the log likelihood in the zero inflated Poisson regression is demonstrated by:
Σyi = 0log (exp (xiβ) + exp (-exp (ziα))) + Σyi<0 (yiziα – exp (zi α)) - Σni= 1log (1+ exp (xiiβ)) - Σyi>0log (yi!) (Lambert, 1992).

Method

Water samples were taken from reservoirs on the Northern Coast and the Bay Island area in Honduras. The Aedes Aegypti that is the primary mosquito transmission vector of the dengue based viruses is derived from an insect that is intimately related with humans and the habitats. The individuals are not only the main source of sustenance for the mosquitos. The reservoirs of water that are present in the residential environments enable the mosquitoes to complete their production. The mosquitos deposit the eggs. The water samples were tested for the presence of mosquito larvae. There were a total of one hundred samples that were derived from samples. This method was conducted by the application of a two by three factorial exam. The variables were the presence or absence of the dengue symptoms. Three groups were reviewed: One group was healthful children who served as a negative control, another group of children that manifested bacterial infection served as a positive control. The final group of children demonstrated serological dengue virus infection with the presence of the IgM and IgG antibodies (Juffrie et al. 2001; Hu et al., 2011). .

Discussion

The dengue infection spreads to approximately twenty million people annually in subtropical and tropical regions. The mortality index that is associated with dengue that is derived from the mosquito vectors is less than two per cent. The range of the disease as it is manifested is extensive. There may be a mild or asymptomatic infection by the diverse extents of the thrombocytopenia and leakage from the vascular system that is a manifestation of the dengue hemorrhagic fever. The extreme cases of dengue involve multi organ failure and severe shock syndrome (Juffrie et al. 2001; Hu et al., 2011; Welch, 1996).
.There is a number of categories of outcome variables that can never be anticipated to fulfill the conventional linear method’s assumption of residuals that are normally distributed. The normally distributed residuals can be found in non- normal outcome variables. There is the requisite of the residuals being measured, unbounded and continuous on a ratio scale or interval. The outcome variables that are categorical do not fulfill this requisite, nor do the count variables. This is less apparent due to their characteristic of being assessed on a proportional skew, which causes their perception to be continuous (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
Variables that have a continuous characteristic assess how much. The variables that have the characteristic of being counted assess how many. The counting variables do not have the quality of being negative numbers, the minimal values are zero. These values are so frequently skewed that zero is the most commonly found value in the majority of the circumstances. These values are also discrete without having the qualities of being continuous (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
The variables that have the capacity of being counted frequently adhere to Poisson or one of its corresponding distributions. In the Poisson distribution process it is assumed that each of the counting variables is the outcome of the similar Poisson process. This is designated as a random process that delineates that each of the events in the counted series is equally probable and independent. In the event that the counted variables is being applied as a result of the regression method, the Poisson regression can be applied on order to provide an estimation of the forecasters influence the quantity of times that the event took place (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
The Poisson method is composed of assumption. One of the assumptions that are disregarded is that the variance is equivalent to the mean. In the event of the variance becoming overly expansive due to the attribute of there being too many zeros in addition to a scarce number of highly rated values, the extension that can administrate the additional variance is the negative binomial method (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
There are times that simply having an excessive number of zeroes that could be forecasted by a Poisson distribution. In these circumstances, a more viable solution is frequently applied as the zero inflated Poisson paradigms. In the event of the excessive variation taking place, the most proximate model that can be applied is the negative binomial paradigm. The zero inflated paradigms operate under the assumption that there are certain zeros that have taken place in a Poisson procedure, and other events did not fulfill the criteria for having the event take place. Consequently, there are two working processes that are functioning. One of the processes defines if the individual event fulfills the criteria for the non-zero reaction; the other function is that which ascertains the counted quality of the response for the individual events (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
The confusion results that both of the process can have the outcome of a zero count. It cannot be determined which of the zeros fulfilled the criteria of eligibility for a non zero count, it cannot be determined which of the zeros were the outcomes of each of the processes. The zero inflation models are applicable in the circumstances of the two regression paradigms. One of the models is a probit or a logistic model that simulates the potential of eligibility for fulfilling a non-zero count. The other model simulates the dimensions of the count. The two paradigms apply the similar forecasting models; however, the coefficients are assessed distinctly. Consequently, the forecasters can have extremely distinct influences on the two processes (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
The application of the zero inflation models has the requisite of some of the individual events being ineligible for the count. Consider the number of field samples in a moth burdock system in an examination of samples derived from reservoirs. There may be some of the samples that could not be considered as vectors of the dengue illness, however the units of observation that is being applied is the moth burdock system. It becomes difficult to conceptualize a situation in which one of the moth burdock systems would possess no potential of vectors that would transmit the dengue illness, notwithstanding that there was no evidence found on some of the field samples in certain circumstances (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
This number can be compared to the number of mosquito stings in a day. This analysis could be adapted to the zero inflation models. There are many of the vectors that simply do not sting subjects and will have an index of zero stings for that particular event. As many of the mosquitoes do not sting respondents in Honduras, there can never have the capacity of manifesting a zero response. The zero inflation models has the character sic of ascertaining which of the forecasters influences the potential of being stung by a mosquito that is a vector of the dengue illness and which of the forecasters can predict how many contacts will be made with the mosquitoes that are vectors of the dengue illness. It is apparent that there may not be identical forecasters in each of the two models or that the forecasters could manifest opposing influences on the two processes (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).
In theory, the count distribution that include the episodes of the outbreaks of dengue illness in Honduras or other events will frequently manifest an anticipated number of observed zeros than the values that are assumed by Poisson distribution. This characteristic is defined as being zero inflated. It may be the case that many of the patients may not be receiving the dengue illness through the vector borne mosquito transmission. This may be the circumstance in the effectiveness trials, where the endeavors that are directed at optimizing the generalization quality of a study by the minimization of exclusionary standards that could place a restriction on the intensity of the challenges at baseline (Juffrie et al. 2001; Hu et al., 2011; Lambert, 1992; Welch, 1996).

Conclusion

The zero inflated models are cost effective in this case due to their consideration of the dual origins of the zero observations. These origins are the sampling zeros that are a component of the foundation sampling distribution and the structural zeroes that do not have the capacity of maintaining any other value besides zero. In the present research proposal it may occur that while some of the respondents acquired scores of zero manifestations of dengue illness due to the lack of vector transmission of the dengue illness, the other respondents had scored zero with regards to being recipients of the dengue illness due to the attribute of having no vectors borne contact with the disease.

References

Juffrie, M., Meer, G. M., Hack, C. E. Haasnoot, K., Sutar, Y. O.,.., and Thijs, L. G. (2001). Inflammatory mediators in dengue virus infection in children: interleukin- 6 and its relation to C- reactive protein and secretory phospholipase A 2. American Journal of Tropical Hygiene, 65(1): 70- 75.
Hu, M-C., Pavlicova, M., and Nunes, E.V. (2011). Zero- inflated and hurdle models of count data with extra zeros: Examples from a high risk intervention trial. American Journal of Drug and Alcohol Abuse, 37(5): 367- 375.
Lambert, D. (1992). Zero inflated Poisson regression with an application to defects in manufacturing. Technometrics, 34(1): 1-14.
Welch, A. H., Cunningham, R.B., Donnelly, C.F. and Lindemayer, D. B. (1996). Modelling the abundance of rare species statistical models for counts with extra zeros. Ecological Modelling, 88: 297- 308.
.Welcome, M. (2013). Roatan health. Roatanet.
**/