Cluster Analysis to detect motivated and non-motivated participants?

[v.codec]

Dear all

I have questions about cluster design and wondering if I can get some advice from you.

I have done a research investigating the correlation between spatial ability and memory capacity. I found there was a moderate correlation between these two variables. When I presented the result in a conference, one of the audience comment that this correlation might simply due to "people who did well in the memory test are people whom are more motivated/serious during the experiment in general, therefore they score as well in the spatial test- and that does not mean these two variables are correlated".

I can understand where he is coming from- but I don't really know how to tackle with this issue. How can I be sure that wasn't the case as the commenter suggested? The commenter suggested that I can use cluster analysis for this type of thing.

I'm not super familiar with cluster analysis and I hope to get some advice from here.
I use SPSS for the cluster analysis.

How do I do this with the cluster analysis? Do I use the memory and spatial scores as the variables and extract the result into two cluster solutions (motivated or demotivated?)

If not, how should I do it? I have been reading about cluster analysis but just couldn't figure out how to solve this issue.

Any advice is appreciated!

 

[Lazar]

Do you have data on motivation or how seriously they took the test? A cluster analysis on memory and spatial scores will most likely just give you four groups high high, high low, low high, low low on menory and spatial scores but nothing on motivation.

Can you give details of your methodology and procedure maybe there is something else you could do if we knew a bit more.

 

[v.codec]

Hi Lazar

We don't have the score for motivation- that's part of the reason I couldn't figure out how to do this.

I used two standard tests for measuring spatial and memory skill. For example, the spatial test requires the participant to select the best picture segment that match the overall picture (it's like logic reasoning test). Similarly for memory test- the participants were presented with a series of number reading to them- and they have to recall the heard number. So the correct number of items are the score of the tests.

 

[Lazar]

And that was it. No treatment conditions or any other data collected? If not I think you are fresh out of luck.

 

[v.codec]

Ref: Lazar

Yes there were other data collected. For instance I have other variables such as auditory skill, music skill A (melody), music skill B (rhythm), music skill C (pitch) so on and so forth...

So all together there were memory, spatial, auditory, 7 types of music skills. It's a within-subjects design (and correlational analysis)

But I don't have the motivation scores.

The general comment was, "how do you know the significant correlation between A and B was not simply due to some participants are just good in general/more motivated etc"...

Does this additional variables help?

 

[Lazar]

hmm maybe you could correlate all of the variables and hope that spatial and memory skill is significant and the other relationships between memory skills and the other variables are not. Arguably motivation should have an effect on all the variables and thus if there are differential relationships (size and sig of correlation) maybe you could argue that it cannot all be motivation.

 

[v.codec]

Thanks Lazar, I also thought of similar thing as you- and I did find the correlations are quite varied between tests

but I just don't understand how cluster analysis would be helpful for this? as he did specifically mention to use cluster analysis...

Did I miss something here?

 

[Lazar]

Im not sure how cluster analysis would help either. Sometime people get enamored and believe it can solve all the worlds ills. This might be the case here.