Applications of Canonical Discriminant Analysis in Horticultural Research

This article reports on soil samples collected from Hsiang-Shan wetland, Taiwan. Canonical discriminant analysis (CDA) was applied to identify an existing habitat type's scheme by identifying the physico-chemical properties of sediment in Hsiang-Shan wetland. The three constructed discriminant functions (CDFs) showed a marked contribution by most of the discriminant variables, and the recognition capacities in these three CDFs were 49.5, 32.8 and 17.7%. Our study revealed that the most important latent factors in Hsiang-Shan wetland are soil texture-caused factor, ocean current-caused factor, nutrient-caused factor, and the redox reaction-caused factor. And the most sensitivity parameters in this habitat followed the descending order: OBD, EC, Eh, sand, TN, porosity, STP, silt, VCP and pH. And the inhabited sediment properties for U. formosensis in terms of soil texture are sand, silt, and clay (34.05, 29.72, and 32.35%, respectively): that is clay loam soil. We also found that U. formosensis preferred to inhabit the upper intertidal zone, spending 8.41% of the time submerged. Vegetation coverage on the ground was less than 2.20%, showing that it preferred to live in a bare intertidal habitat. Concerning nest choosing, excavating burrows is more difficult when a high soil penetration force is required, and in this study the soil penetration force for 20 cm was found to be is 45.98 N/cm(2). The results will be helpful in developing a methodology for use by the government in refining its management programs.

Section: Methodsmentioning

confidence: 99%

Status and habitat preferences for endemic inhabitants of fiddler crab Uca formosensis in Hsiang-Shan wetland, Taiwan

Liao¹,

Chang

Lin

2007

“…Total standardized canonical coefficients specify the joint effects of independent variables of a given CDF, so are more informative than the total canonical structure coefficients (Rencher, 1992). Total standardized canonical coefficients can be misleading when independent variables are related (Cruze-Castillo et al, 1994). In this work, the total canonical structure coefficients were used to interpret the CDFs because significant correlations obtained among some of the independent variables (Table III), and the total standardized canonical coefficients was used to yield the CDFs.…”

Section: Discriminant Analysismentioning

confidence: 99%

Water Quality during Development and Apportionment of Pollution from Rivers in Tapeng Lagoon, Taiwan

Liao

Lai

Chen

et al. 2006

This work investigated water samples collected from Tapeng Lagoon, Taiwan. Factor analysis was conducted to explain the characteristics and the variation in the quality of water during the disassembly of oyster frames and fishery boxes. The result shows that the most important latent factors in Tapeng Lagoon are the ocean factor, primary productivity factor, and the fishery pollution factor. Canonical discriminant analysis is applied to identify the source of pollution in neighboring rivers outside Tapeng Lagoon. The two constructed discriminant functions (CDFs) showed markedly contribution to all discriminant variables, and that total nitrogen, algae, dissolved oxygen and total phosphate, combined in the nutrient effect factor. The recognition capacities in these two CDFs were 95.6%, 4.4%, respectively. The water quality in the Kaoping river most strongly determined that in the Tapeng Lagoon the best is. And disassembling the oyster frames and fishery boxes improves the water quality markedly. However, environmental topographic conditions indicate that strengthening stream pollution prevention and to constructing another entrance to the ocean are the best approaches for improving the quality of water in Tapeng Lagoon, especially by reducing eutrophication. These approaches and results yield useful information concerning habitat recovery and water resource management.

“…Two main statistics explain the properties and structure of the canonical discriminant functions -(1) the total canonical structure coefficients that indicates correlation coefficients between individual variables and canonical scores (similar to variable loadings in factor analysis) and (2) total standardized canonical coefficients are multipliers of standardized independent variables and produce standardized canonical scores. The validity and the use of these two indices have been discussed but controversies still remain (Rencher, 1992;Huberty, 1994;Matthew et al, 1994;Cruze-Castillo et al, 1994;Momen and Zehr, 1998). Total standardized canonical coefficients reflect the joint effects of independent variables of a given CDF, and thus, are more informative than the total canonical structure coefficients (Rencher, 1992).…”

Section: Discriminant Analysismentioning

confidence: 99%

“…Total standardized canonical coefficients reflect the joint effects of independent variables of a given CDF, and thus, are more informative than the total canonical structure coefficients (Rencher, 1992). Whereas, total standardized canonical coefficients can be misleading when independent variables are interrelated (Cruze-Castillo et al, 1994). In this work, the total canonical structure coefficients was used to interpret the CDFs because of significant correlations among some of the independent variables (Table I), and the total standardized canonical coefficients was used to construct the CDFs.…”

Section: Discriminant Analysismentioning

confidence: 99%

“…Factor analysis can determine the most important factors that contribute to structure of the data (Borovec, 1996;Gupta et al, 1999;Soares et al, 1999;Jenerette et al, 2002). Canonical discriminant analysis is used interpret spatial distribution of bioassemblage with various environmental parameters (Cruze-Castillo et al, 1994;Mitchell et al, 1997;Momen and Zehr, 1998;Ogutu, 1999;Shin and Fong, 1999). In this work, first we analyzed soil properties in the Kuan-Tu wetlands.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Interpretation and Discrimination of Marshy Wetlands by Soil Factors in the Kuan-Tu Natural Park, Taiwan

Liao¹,

Chang

2005

This work investigated soil samples collected from Kuan-Tu wetlands, Taiwan. Factor analysis was performed to explain the impact of various soil factors on this marshy wetlands located in suburban Taipei. The results indicated that the latent factors were heavy metals, salinity, and soil organic matter. Canonical discriminant analysis was used to improve an existing vegetation classification scheme by identifying the physical-chemical properties of sediment in Kaun-Tu wetlands, Taiwan. Predictive discriminant analysis was used to examine the ability of the models to predict class membership for unknown soil sample. Multivariate analysis of the spatial patterns of soil quality and vegetation types showed that different properties of soil grew different types of vegetation and absorbed contaminants differently. We can feasibly conserve a suitable habitat for wetland biology by processing these unstable predictor variables. The methodology and results provide useful information concerning the Kuan-Tu wetlands and may be applicable to other wetlands with similar properties that are experiencing similar environmental issues.