SensMixed (version 2.1-0)

sensmixed: Analysis of sensory data within a mixed effects model framework

Description

Constructs mixed effects models for each of the selected by a user attributes. By default the largest possible models (that contain all possible interactions in fixed and random parts) are fitted. The complexity of the fitted models can be changed. Non-significant random effects are eliminated (by default). The likelihood ratio test (LRT) is used for testing the random terms, F-type hypothesis test is used for testing the terms. The type of the model and the type of the analysis can be changed with the control argument (see sensmixedControl())

Usage

sensmixed(attributes, prod_effects, assessor, 
          replication = NULL, data, product_structure = 3,
          error_structure ="ASS-REP", MAM = TRUE,
          control = sensmixedControl())

Arguments

attributes

a vector with names of sensory attributes

prod_effects

a vector with the names of the variables related to the product

replication

a character with the name of the replication column in the data, if present

assessor

a characthe with the name of the column in the data that represents assessors

data

a data frame (data from sensory studies)

product_structure

numeric, takes values in in c(1, 2, 3). Specifies the complexity of the fixed part (product effects) of the mixed effects models for all attributes.

product_structure = 1

only main effects

product_structure = 2

main effects and 2-way interactions

product_structure = 3

all main effects and all possible interaction

error_structure

character, takes values in c("ONLY-ASS", "ASS-REP"). Specifies the complexity of the random part of the mixed effects models for all attributes.

error_structure = "ONLY-ASS"

assessor effect and all possible interactions between assessor and product effects

error_structure = "ASS-REP"

assessor and replicate effect and interaction between them and interaction between them and product effects

MAM

logical. if TRUE then mixed assessor models (MAM) are fitted for the selected attributes (see Brockhoff, P. B., Schlich, P., & Skovgaard, I. (2015))

control

a list (of class sensmixedControl()) containing control parameters. See the sensmixedControl() documentation for details.

Value

FCHi

matrix with Chi square values from LRT test and F values form F-type test for the selected attributes

pvalue

matrix withp-values for all effects for the selected attributes

References

Brockhoff, P. B., Schlich, P., & Skovgaard, I. (2015). Taking individual scaling differences into account by analyzing profile data with the mixed assessor model. Food Quality and Preference, 39 , 156-166.

Kuznetsova, A., Christensen, R. H., Bavay, C., & Brockhoff, P. B. (2015). Automated mixed ANOVA modeling of sensory and consumer data. Food Quality and Preference, 40, Part A, 31 38. URL: http://www.sciencedirect.com/science/ article/pii/S0950329314001724. doi:http://dx.doi.org/10.1016/j. foodqual.2014.08.004.

See Also

sensmixedControl, conjoint, SensMixedUI

Examples

# NOT RUN {

## import SensMixed package
library(SensMixed)

## convert some variables to factors in TVbo
TVbo <- convertToFactors(TVbo, c("Assessor", "Repeat", "Picture"))

## run automated selection process
res <- sensmixed(c("Coloursaturation", "Colourbalance"),
prod_effects = c("TVset", "Picture"), 
assessor="Assessor", data=TVbo, MAM=TRUE)

res


## run MAManalysis function
res_MAM <- sensmixed(c("Coloursaturation", "Colourbalance"),
                 prod_effects=c("TVset"), replication="Repeat", 
                 assessor="Assessor", data=TVbo, control = list(MAM_balanced=TRUE))
## print is not yet implemented
## get anova part
res_MAM[[3]][,,1]

## compare with the general implementation
res <- sensmixed(c("Coloursaturation", "Colourbalance"),
                  prod_effects=c("TVset"), 
                  assessor="Assessor", data=TVbo, MAM=TRUE, 
                  control = list(reduce.random=FALSE))
res$fixed      

# }
# NOT RUN {
res <- sensmixed(names(TVbo)[5:(ncol(TVbo) - 1)],
                  prod_effects=c("TVset", "Picture"), 
                  assessor="Assessor", 
                  data=TVbo)


plot F and Chi square values
plot(res)
# }