I am using mixture optimal design( RSM ) for composite flour .One of my response model is non significant. what does it mean?And for the same response no coded equation is given by the software? The response is foaming stability.
Response Surface Methodology (RSM) performs a multivariate polynomial regression of second order. The equation produced is a regression model which is used to predict the target/output on the basis of input variables. An non-significant model means that the regression model was not able to accurately predict the outputs on the basis of responses.
Let us assume you want to predict foaming stability on the basis of two independent variables (X1 and X1). The regression model will be as following:
The insignificant of the response model mean that the experimental model is not accurate in predicting the experimental response. The insignificant of the model could be caused by many things. First, it could be caused error result from the experimental design. it always advice that the experimental data are randomized to minimized error. Another reason could be your experimental data or response obtained from the experimental design contradict the assumption of analysis of variance. Once a data contradict the assumption of variance, such data need to be transformed. In the case of response, it doesn't contradict the assumption of ANOVA since your response (foaming stability) is a measured variables. However, your response might need to be transformed if the ratio between the highest value (foaming stability) and the lowest foaming stability is greater than or equal to 5. Try and check the ratio to see if its not greater than 5. Another way to check what might be wrong with your experimental data or design is to check the diagnostic tool (This depends on the software but can easily be seen if you are using Design Expert of Statease). From the Diagnostic tool, check if your data are evenly distributed to follow normal probability distribution. Also check the Outlier t Plot which is a plot of the residuals against experimental runs, a value above the threshold line indicate that the experiment of that run needs to be performed again.
Wish I could take you through how you can solve the problem but kindly go through the handout attached below.