A Particular Place of Residence of a Person - Research Paper Example

? Geodem Introduction The concept of area ification is indispensable in modern demographic dis Its relevance stems from the various uses of most population data and the need to efficiently allocate resources in tandem with population parameters. The paper seeks to give a transparent illustration of the methodologies that are used in creating the common Output Area (OA) classification. This is done through a dissection of the cluster analysis and the way of the area classification. Several principles actually support the decision to select the given variables and standardization methods. Indeed, one of the ways of solving large volumes of information and mitigating their complexity is to understand the means of simplifying such information into comprehensible terms. This is achieved through the development of area classifications and the creation of composite indicators. Area classification Area classification is the categorization of particular areas on the basis of similarities and commonalities that exist. These include features and other natural or artificial characteristics. One of the commonly used classifications is the geo-demographic classifications. This is the analysis of populations depending on where they stay. Geo-demographics is hinged on the inevitable assumption that population and place are inextricably tied together. Area classification creates a peculiar way of grouping together spatial patterns from a given range of variables and finding the similarities and differences between them. A scheme of classification applied is therefore a representation of the technique for the arrangement of data in order to improve the recovery of information and efficiency. For the 2001 census in the UK, the OAs are actually the smallest demographic units for which the data is released. The census agencies for England, Scotland, and Northern Ireland were responsible for the creation of the particular outputs for the three countries. Although there were some differences in the methodology between the three agencies, the general output actually reflected a matter of cross-sectional analysis. In Northern Ireland, the agency followed the ONS methodology, which had a minimum Output Area of 40 households and 100 residents. However, in Scotland, most of the OAs reflected the data from the 1991 census which had the figure of 20 households and 50residents Cluster analysis The process of area classification is done through the clustering or grouping of geographical units by the use of particular methods. Cluster analysis borrows largely from clustering algorithms but is instead much more than the mere grouping of the objects. Therefore, to successfully run a cluster analysis will require a series of particular steps, which involve multiple decisions across all the stages. At this point, there is no wrong or right answer to most of the decisions that are made in the process. Every decision in the stages will therefore give its own alternative results. As a result, the different decisions gathered could be appropriate or not appropriate depending on the classification that is to be created. There are seven steps involved a successful running of the cluster analysis. Each step in the framework represents a very important decision point that is imperative for the smooth running of the cluster analysis (Harris and Webber, 2005). It is suggested that the user of the method should be in a position to correctly recognize all the important and critical decisions involved and their individual influences on the results (Everett and Leese, 2001). At this point, it becomes important to clearly distinguish between the conventional cluster analysis and the clustering method. Clustering method connotes the simple process by which the clusters are actually formed (Everett and Leese, 2001). On the other hand, cluster analysis is much elaborate and implies the wider series of steps that have to be followed in order to finish the whole analysis. It is normally essential for the users of cluster analysis to note down and report all the decisions that are made at each step of the process. This becomes more important if the classification is to be done by other people and not the users. The seven steps involved in this classification are outlined below. Step 1: Clustering of the elements It is suggested that these clusters should, if possible, be able to give a total geographical representation and that the cluster structure that is believed to be present should be represented by these clusters. Further, all the clusters should be sampled well in case the generalization to a wider population is needed. Step 2: Clustering of variables (these are the attributes of the objects) Here, the variables must represent all the measurements that are taken on each area that is to be grouped. Variables are only included if there is sufficient justification for their presence. Variables that can hide significant patterns across the variables must be avoided. Such variables include irrelevant or masking variables. Step 3: Variable standardization In this criterion, standardization is not really a requirement and can therefore be ignored in any set of data. The researcher is at his discretion to decide whether to apply standardization or not. If yes, he has to decide the particular standardization method to use. Step 4: The proximity measure At this point, a particular measure of the similarity or dissimilarity has to be identified in accordance with the clustering variables. This will show the degree of separation or closeness in the objects to be grouped or clustered. Different methods can be applied in this case. For instance, the Euclidean distance is a good measure of dissimilarity as it reports bigger values as the two entities in comparison become less dissimilar. On the contrary, the measure of similarity will report larger values as the two entities become similar. There is no restriction on the measures to be used. Both linear and non-linear apply. Step 5: The clustering method The use of this method requires that all the steps that are used should be those that are customized to recover the clusters that are suspected or thought to be present. This method is very important since different clustering methods are efficient for given types of cluster structures. The method selected should be very robust and should manage to handle a variety of data amounts. Step 6: decision on the number of clusters This is one of the most difficult and yet critical decisions to be made in the cluster analysis. It becomes more complex incase no existing information is available on the number of clusters that are supposed to be found on the data. However, there are several rules that govern the selection process, which in most cases create many contradictions. There is no definite answer on the number of clusters to be selected, as the decision is never based on scientific theory. The answer selected should always be governed by its usefulness and not the correctness of its representation. Step 7: Interpretation of the results and testing The last step involves the interpretation of the results as an assessment of whether it’s a solution to the problem at hand. This normally requires sufficient knowledge on the particular subject matter of the question. Standardization of data All the techniques of clustering were set on either the similarity or the dissimilarity of the particular cases (Kaufman and Rousseeuw, 2005). This was gauged by forming a distance matrix that reflected all the particular variables in the data set. Problems become inevitable if the scales or magnitudes are quite different among the variables. The variables that contain larger dispersion have much influence on the ultimate measure of similarity. It was therefore imperative to carefully represent all variables by equal measure in the distance by the standardization of the initial data. The methods of standardization that were considered include; Z-score standardization, range standardization and the internecine range standardization. Defining similarity When every variable has been correctly standardized, it is important how near or far the cases be to each other. Most of the techniques used in cluster analysis normally start with a matrix which quantitatively reflects a measure of the similarity between every pair of the data to be classified. To be ‘close’, a pair of cases must have a small distance between them and a big similarity. Different methods are used to quantify distance which include Ward’s method and the Euclidean distance. Classification of the statistical wards The 2001 classification of wards involved the adoption of an approach recommended by the Advisory Board. The procedure involved a series of steps below; It began by the generation of a random classification of the clusters into wards of 1000. The k-means method was then used together with the original cluster centers as the starting point in order to reach the full 1000-clusters. At this step, Ward’s method was used for the subsequent 1000 clusters from the k-means. The number of categories and sub-categories is determined through the analysis of the agglomeration schedule. The subgroups that came from the Ward’s model are further refined by the use of k-means so that each ward is assigned the relevant subgroup. The rest of the levels are found through the hierarchy from the Ward’s method. For convenience, the original number of the clusters in the first step should be very large. The selection of the number of clusters is important but does not nevertheless influence the result. The basic tenet that governs area classification is that people who stay close to one another have the propensity to relay similar characteristics (Martin, 2002). This is referred to as spatial autocorrelation, which is based on the principle that similar phenomena are found close to one another. It is imperative to realize that people’s lives are normally pegged on particular places, which include the place of birth or where the people have lived. Such placers can be as large as cities and as small as neighborhoods. A particular place of residence of a person will, in a way, influence their identity. In conclusion, it can be stated that the classification can be effectively used to account for the processes and patterns within particular subject areas. Such areas include the general elections. The whole process that was used by the agencies was actually effective and appropriately served the demands of time. References Everitt, BS Landau and Leese M 2001, Cluster Analysis, Edward Arnold, London. Everitt, BS, Landau, S and Leese, M 2001, Cluster Analysis 4th Ed., Edward Arnold, London. Harris, R, Sleight, P and Webber, R 2005, Geodemographics, GIS and Neighbourhood Targeting, Wiley, Chichester. Kaufman, L and Rousseeuw, PJ 2005, Finding groups in data, Wiley, Chichester. Martin, D 2002b, Output Areas for 2001, Chapter 3, pp, 37-46 in Rees, P, Martin, D and Williamson,P (Eds.), The Census Data System, Wiley, Chichester. Read More