Newbie question: Hierarchical clustering - test value (Lebart, 2000)

KNIME Analytics Platform
1

Hello,
 
I am new to KNIME and I previously used Tanagra to do my clustering.
Tanagra has a nice feature called "Group characterization". Basicaly what it does
is
1- calculate the means and standard deviations within each group for each of the columns used
2- calculate mean and standard deviations of the full the dataset for each column
3- calculate a "test value" withing each group for each of the columns used
 
I was able to perform (1) and (2) (1 = "Statistics > Statistics" and 2 = "Data Manipulation > Row > Transform > Group By")
 
But I have no clue to how I could perform (3). The "test value" calculated by Tanagra is defined as follows (Lebart et al. 2000; p.181)
 
test value for a specific group and column =
( group mean - dataset mean ) / Square root of ( ( ( total number of rows - number of rows in group ) / ( total number of rows - 1 ) ) * ( variance of the dataset / number of rows in group ) )

http://eric.univ-lyon2.fr/~ricco/tanagra/fichiers/en_Tanagra_Comprendre_La_Valeur_Test.pdf
 
How could I calculate such a "test value" for each column in each group of the resulting clusters ?
 
Thanks in advance for your answers.

2

Sounds like you are after something like a single sample t test. The formula you mention is rather similar.

this is available under statistics and then hypothesis testing.

does this help.

 

simon.

3