Help for standard deviations K-means clustering

Comments

• koenderks

  February 2022

  Hi Manon,

  This can be achieved by exporting the clusters to your data set by clicking “Add predictions to data” and filling in a name for the new column. Then, you can go to the descriptives analysis and use this new column to split the data (i.e., by dragging this variable to the “Split” box). Then, you should get the descriptives (mean, sd, etc.) for all clusters.

  Koen
• FrenchOrtho

  February 2022 edited February 2022

  Hello,

  it doesn't work. indeed I have done the instructions you gave me but it gives me only one table with one average while I have in my analysis three different clusters.

  Moreover, I put my analysis in the variables and not in split because otherwise I have no results in the table...

  I am attaching the examples in images

  thank you

  
  

  

  

  
  

  

  
  
• koenderks

  February 2022

  Hi Manon,

  If the cluster memberships are in the variable "macro VS micro" then this variable should be dragged to the "Split" box. You should then insert all the variables you have used in your initial cluster analysis in the box under "Variables".

  Say you have used the variables "A", "B", and "C" to find the clusters. You can then export these memberships to a new variable "Z". In the descriptives analysis you should then drag the variables "A", "B", and "C" to "Variables" and the variable "Z" to "Split".

  Let me know if it works!
• FrenchOrtho

  February 2022

  Hello,

  Here are the results I get, it does not correspond to the results I have already noted...

  I must admit that I'm starting to despair, the software that seemed simple to use is finally complicated, I don't find the same results each time, they are different depending on the table or the graphs, as well as the names of the clusters, I feel really lost and I feel like I can't move forward.... because I don't have an answer to all my questions asked on the forum.

  
  

  

  

  
  
• koenderks

  February 2022 edited February 2022

  Hi Manon,

  Can you try the the following steps:

  1. Enter your raw data (no z-scores) in the "Variables" box in the cluster analysis.
  2. Go to "Advanced options" and make sure "scale variables" is off.
  3. Click to box "Set seed" to get reproducible results
  4. Now export your clusters to your data.
  5. Now go to descriptives and make the table (enter the raw data!) split on the cluster membership, the cluster means should now match. Your discrepancy is because in the cluster analysis the data is automatically scaled to a z-score, but that means that the cluster means are also on the z-scale.
  6. I would not look at the t-sne plot for any meaningful information, just go to descriptives --> Plots --> Scatter plots and it will automatically color the points according to cluster membership.
  

  
  
• FrenchOrtho

  February 2022 edited February 2022

  Hi!

  
  

  After following your instructions, this is what I have, I'm sorry but it doesn't work... For the microstructure Z-score, I was forced to create this score because my microstructure score contains 3 different data with different scales, so I can't average the three to get a raw score. Moreover by checking "set seed", I have too many clusters...

  Again, even with your directions, I can't find the same results in the tables.... and the graph doesn't even open :'(

  Also, I would need to know which point corresponds to which participant, is that possible?

  
  

  Thank you for your time,

  
  

  Manon