BookmarkSubscribeRSS Feed
Choose Language Hide Translation Bar

Community Trekker


Oct 27, 2015

DoE for inter laboratory comparison

Dear all,

I need to define a common test plan for a number of different laboratories A,B,C,...J that use two types of specimens (Type 1 and Type 2). Let's say labs A - D can only use type 1 and labs E - J can only use type 2 and every lab can do 4 tests. Let's consider only 1 additional factor "strain". I want to include the categorical factors "Lab" and "Specimen Type" in the test plan to check whether there are significant impacts from either of them (hopefully not).

In order to match the labs and the specimen type I have created a table containing the labs in the fist column and the type of specimen in the second column. Since every lab can do for tests, the first 8 entries in the first column are A A A A B B B B... and in the second column Type 1 Type 1 Type 1 Type 1 Type 1 Type 1 Type 1 Type 1... I'm using this table as covariate matrix (see attachment).

I use the DoE custom designer for making the test matrix. I add the two covariate factors from the table and an additional continuous variable strain. The model should contain only the main effects. When I run the Design Generation, I get an error message "The selected terms in the Model outline are linearly dependent on the previous terms. Please press the Remove Term button and try again." The selected term in the model is "Specimen Type". I think the problem is that the factors "Lab" and "Specimen Type" vary simultaneously (every lab can only test a single type of specimen).

However, since four labs use the same specimen it should be possible estimate whether "Lab" and "Specimen Type" have a significant impact or not. How should I approach this problem?