BookmarkSubscribeRSS Feed
Choose Language Hide Translation Bar
pratyushdash
Occasional Contributor

JMP 12 and JMP 14 K means cluster difference

In JMP 12(left picture below) for K-means clustering we had a label option. I dont seem to find the same option on JMP14(right pic below). How do i get to do the same thing without the label option? I tried running just by adding the Y columns but the results are varying. Please help!

JMP Question.png

0 Kudos
2 ACCEPTED SOLUTIONS

Accepted Solutions
gzmorgan0
Super User

Re: JMP 12 and JMP 14 K means cluster difference

Hmmm, I have JMP PRO v12.2 (as well as 14.2) and from the picture below you can see that my version of JMP 12.2 does not have a Labels role. From my experience, the variable you select for Label is used in Hierarchical clustering. If no Label is specified, then the row numbers are used.  See pic 2 & 3.  You can always select the column (sex) then from the Main Menu >Cols, or from your table's Columns red menu, select Label.  Then the kMeans cluster graphs you can highlight points and see their locations. 

 

Alternately, you can save your cluster numbers and analyze the relationship of your Label variable versus the Cluster ID. I am not sure of your goal, so I am just documenting my version of JMP 12 and my experience. 

 

image.png

image.pngLabel variable is seximage.pngNo Label variable is specified

 

eclaassen
Staff

Re: JMP 12 and JMP 14 K means cluster difference

As @gzmorgan0 noted, labels are only used in Hierarchical clustering. They are not used in KMeans. Though you could previously specify labels, they were not used in a KMeans analysis. If you are obtaining different results, it is likely because the algorithm uses random cluster centers to begin the clustering. This can result in slightly different clusters in some cases.

For a case where this does not occur, try using the Iris.jmp data table. Using the older version, you can add "Species" as the label with the remaining columns as Ys. Then in the current version specify the Ys. The results (on my machine) are identical.

4 REPLIES 4
gzmorgan0
Super User

Re: JMP 12 and JMP 14 K means cluster difference

Hmmm, I have JMP PRO v12.2 (as well as 14.2) and from the picture below you can see that my version of JMP 12.2 does not have a Labels role. From my experience, the variable you select for Label is used in Hierarchical clustering. If no Label is specified, then the row numbers are used.  See pic 2 & 3.  You can always select the column (sex) then from the Main Menu >Cols, or from your table's Columns red menu, select Label.  Then the kMeans cluster graphs you can highlight points and see their locations. 

 

Alternately, you can save your cluster numbers and analyze the relationship of your Label variable versus the Cluster ID. I am not sure of your goal, so I am just documenting my version of JMP 12 and my experience. 

 

image.png

image.pngLabel variable is seximage.pngNo Label variable is specified

 

pratyushdash
Occasional Contributor

Re: JMP 12 and JMP 14 K means cluster difference

Thank you so much for your help!
0 Kudos
eclaassen
Staff

Re: JMP 12 and JMP 14 K means cluster difference

As @gzmorgan0 noted, labels are only used in Hierarchical clustering. They are not used in KMeans. Though you could previously specify labels, they were not used in a KMeans analysis. If you are obtaining different results, it is likely because the algorithm uses random cluster centers to begin the clustering. This can result in slightly different clusters in some cases.

For a case where this does not occur, try using the Iris.jmp data table. Using the older version, you can add "Species" as the label with the remaining columns as Ys. Then in the current version specify the Ys. The results (on my machine) are identical.

pratyushdash
Occasional Contributor

Re: JMP 12 and JMP 14 K means cluster difference

Thanks a lot for your help. Appreciate it!
0 Kudos