Additivity and interaction in three-component experiments with mixtures (Q3678508)

From MaRDI portal

!

This is the item page for this Wikibase entity, intended for internal use and editing purposes. Please use the normal view instead:

scientific article; zbMATH DE number 3899986
Language Label Description Also known as
default for all languages
No label defined
    English
    Additivity and interaction in three-component experiments with mixtures
    scientific article; zbMATH DE number 3899986

      Statements

      Additivity and interaction in three-component experiments with mixtures (English)
      0 references
      0 references
      1985
      0 references
      linear blending model
      0 references
      synergism of the mixture
      0 references
      additive model
      0 references
      blandness
      0 references
      interaction
      0 references
      polynomial approach
      0 references
      mixture experiments
      0 references
      canonical forms
      0 references
      factorial approach
      0 references
      In an experiment with mixtures the distribution of the response variable depends on the proportion of the components of a mixture and not on the absolute amounts of those components. In this paper it is assumed that the mixture has three components. Their proportions are denoted by \(x, y\), and \(z\) and interest is centered on the way in which its expected value \(\eta(x,y,z)\) depends on \((x,y,z)\). The simplest model that can be assumed for \(\eta\) is the linear blending model LB defined by LB: \(\eta (x,y,z)=\lambda x+\mu y+\nu z\), for which it follows immediately that \(\lambda =\eta (1,0,0)\), \(\mu =\eta (0,1,0)\), \(\nu =\eta (0,0,1)\). The difference \(\eta(x,y,z)-x\eta (1,0,0)- y\eta (0,1,0)-z\eta (0,0,1)\) is called the synergism of the mixture \((x,y,z)\). The additive model AM is defined by AM: \(\eta (x,y,z)=\alpha (x)+\beta (y)+\gamma (z)\) for some functions \(\alpha\), \(\beta\), \(\gamma\). Having made \(\alpha\), \(\beta\), \(\gamma\) unique, write \(\alpha (x)=\lambda x+\theta (x)\), where \(\theta (1)=0\) as well as \(\theta (0)=0\), with corresponding expressions for \(\beta (y)\), \(\gamma (z)\). The net result is that if AM is expressed as AM: \(\eta (x,y,z)=\lambda x+\mu y+\nu z+\theta (x)+\phi (y)+\psi (z)\), where \(\theta (0)=\theta (1)=\phi (0)=\phi (1)=\psi (0)=\psi (1)=0\), then the constants \(\lambda\), \(\mu\), \(\nu\) and the functions \(\theta\), \(\phi\), \(\psi\) are uniquely defined by \(\eta\). If \(\eta\) satisfies AM and \(\psi\) is identically zero then, on writing \(z=1-x-y\), \(\eta\) can be expressed in the form \(\eta (x,y,z)=\alpha^*(x)+\beta^*(y)\). In this case we shall call the third component bland. Additivity and blandness of all three components is equivalent to linear blending. Given AM, the synergism of the mixture is the sum \(\theta (x)+\phi (y)+\psi (z)\) of the three nonblandnesses. If \(\eta\) does not satisfy AM, so that there is nonadditivity in \(\eta\), we shall say that there is interaction between the three components. In this paper, the polynomial approach to mixture experiments is examined in the light of AM. This leads to new canonical forms for polynomial models. Also, a factorial approach is developed. Finally, there is a brief discussion of other approaches, unrelated to AM, to mixture experiments. A numerical example is given.
      0 references
      0 references
      0 references

      Identifiers