Search the Elsmar Cove!
**Search ALL of Elsmar.com** with DuckDuckGo including content not in the forum - Search results with No ads.

Data analysis Design of Experiments

Movic

Registered
#1
Hello,

I have a full factorial 2^4 Design of Experiments for sauce analysis, so I have made 16 samples in total. Of each sample I made one large batch, so no replication.
These 16 samples were then subjected to a sensory panel consisting of 16 members. They all scored all samples on several sensory attributes twice, so for each sensory attribute I have 32 scores per sample.

My question now is, how do I put this in Minitab? I know that the samples itself are not made in replicates, because I made a batch of each, but the replication in analysis confuses me. Should I treat each panel member AND the repetitive tasting as a repeated measure, and take the average of all 32 scores per sample for analysis (so in total I will have 16 rows for 16 samples)? Or should I include extra rows?

Thanks!
 
#2
Hello,
My question now is, how do I put this in Minitab? I know that the samples itself are not made in replicates, because I made a batch of each, but the replication in analysis confuses me. Should I treat each panel member AND the repetitive tasting as a repeated measure, and take the average of all 32 scores per sample for analysis (so in total I will have 16 rows for 16 samples)? Or should I include extra rows?
Thanks!
I would have assumed and ensured that
* samples are coded ( panel members are blinded)
* at-least triplicate measurements.
* each measurement carried out on same day or different days etc
* measurements carried out over a period of time. ( if that is relevat)
etc

AND solution could depend upon, analysis of each panel-member rating of each sample wrt to determine if there is bias of a
* particular experiment leading to wider/broader range of assessment
* particular sample having a wider/broader range of assessment

in short,
I would add rows for assessment of each panelist ( if inter-panelist assessment is coherent) (==> 16 experiments x 16 panelists x 2 measurements each )
i would translate the assessment of sample by the ALL-panelists into a representative value ( viz., std deviation, average/median, etc) ( ==>16 experiments, multiple responses)
 
#3
I would have assumed and ensured that
* samples are coded ( panel members are blinded)
* at-least triplicate measurements.
* each measurement carried out on same day or different days etc
* measurements carried out over a period of time. ( if that is relevat)
etc

AND solution could depend upon, analysis of each panel-member rating of each sample wrt to determine if there is bias of a
* particular experiment leading to wider/broader range of assessment
* particular sample having a wider/broader range of assessment

in short,
I would add rows for assessment of each panelist ( if inter-panelist assessment is coherent) (==> 16 experiments x 16 panelists x 2 measurements each )
i would translate the assessment of sample by the ALL-panelists into a representative value ( viz., std deviation, average/median, etc) ( ==>16 experiments, multiple responses)
Thanks for your reply. If I understand correctly, you would
1) Add values from all panelists and all measurements separately in Minitab
2) Calculate the average, standard deviation, median, etc for each sample
3) Perform further analysis, such as ANOVA, analysis of factorial design on these averages?

Or would you also use all data to perform these analyses?
 
#4
Or would you also use all data to perform these analyses?
allow me to rephrase,
if within a given sample, variation among panelists is acceptable, then consider the average -std dev - range as responses. ( 16 experimens, and multiple responses)

otherwise, consider all responses as rows. ( 16x16x2 measurements )
 
Top Bottom