A carregar...
Publicação
Multivariate data analysis for monitoring the quality of the commercialized bottled water in Bangladesh
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.5 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Several multivariate statistical or chemometrics or pattern recognition techniques e.g. Principal
Component Analysis, Factor Analysis, Hierarchical and Non-Hierarchical k-Mean Cluster Analysis
have been applied to gain understanding about the quality of the packaged bottled drinking
water in the market of Bangladesh. Twenty three (23) physico-chemical properties of total of 51
water samples have been investigated. The data set consists of 49 individuals from 11 Brands
and 2 deionized ASTM TYPE-I water samples produced in the laboratory to be a technically pure
water having Electrical Conductivity ~0.056 μS-cm-1. Descriptive statistics, analysis of variance,
Non-Parametric Kruskal-Wallis tests have been conducted to detect statistical differences
between the water types and different brands. Total of 23 attributes of water covering major
ion contents: sodium, potassium, calcium, magnesium, iron, manganese, chloride, fluoride,
sulphate, bicarbonate and nitrate and other features: pH, temperature, total dissolved solids,
electrical conductivity, hardness, ammonium, nitrite, free cyanogen and chemical oxygen
demand, total cation sum and total anion sum. Both the Principal Components Analysis and the
Factor Analysis revealed that the differences between water individuals are best characterized
by four Principal Components or Factors indicating material loadings, hardness or softness
aesthetic acceptability and lightness/sutability for human consumption. Hierarchical and Non-
Hierarchical k-means Cluster Analysis clearly identified the presence of four distinct clusters: A,
B, C and D among the bottled water products in the market of Bangladesh. The profile features
for each cluster have been defined as such the classification achieved to acquire improved and
detailed understanding of the general properties of the products under study. We have
observed that HCA using WARD algorithm provided us with more realistic classification solution
in comparison with non-hierachical k-means as the Cluster members are truly reflecting their
group pattern in line with their chemical compositions. HCA using WARD showed that BRAND05
and BRAND11 belonging to Cluster A products execssively loaded with materials and considered
to be as hard waters. And BRAND09 and BRAND10 staying with DEIONIZEDWATER belonging to
Cluster B are completely devoide of essential minerals as such seemed to be as ultra low mineral
content type water or too soft in nature. The other folks BRAND03, BRAND04, BRAND06,
BRAND07 and BRAND08 are also not having sufficient mineral contents so as to be very soft
water indeed. Hence, waters belonging to Clusters A, B and C are not suitable for human
consumption. Only two brands BRAND01 and BRAND02 staying in Cluster D appeared to be
suitable for human consumption in every respect.The fact is the BRAND01 is produced by a
foreign manufacturer. That means, all other local brands, except BRAND02 are essentially not
having the appropriate quality to be drinking waters. From both PCA and FA these two brands
BRAND01 and BRAND02 have been very well explained. These are the major outcomes of this
study not immediately apparent from univariate approach or not appeared from the data set
while looking through naked eyes. It is revealed that the multivariate data analytical techniques
have potential to be useful complementary techniques to support the existing univariate
practices for industrial quality assurance quality control, market surveillance, standardization
process and or regulatory purposes and also seemed to be interesting to academic and scientific
communities seeking advanced knowledge.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Statistics and Information Management, specialization in Information Analysis and Management
Palavras-chave
Chemometrics Pattern Recognition Principal Component Analysis Factor Analysis Cluster Analysis Bottled Water Quality Monitoring Market Surveillance