Complex correlations

Why is 42 in b and c? I don't think I follow this...

I know it's frustratingly complicated sorry; that's why I want to automate it as I keep making mistakes manually:
the difference between 42 and 58 is less than 17 so they have to be in the same group,
The difference between 42 and 35 is less than 17 so they have to be in the same group,
but,
The difference between 35 and 58 is more than 17 so they have to be in different groups,
Therefore 35 and 58 are in different groups but 42 is has to be in both groups.

Thanks to anyone spending time on this mind bender!

OK, I see. How many data items are you likely to have - is seven just illustrative?

seven is the max it's usually 'only' five.

This works for your example, but better test it thoroughly.

Please Login or Register  to view this content.

Thats wonderful thank you! I'm going to test it out here with a few different sets.
I'm very grateful you know =)

Thanks Stephen,

This is getting really close it's great!
I've attached a sheet with a few examples of working and not working groups, your latest code from yesterday is in module 1, I just changed nLSD to relate to cell D2 as it's value differs depending on the variability between means.

I always wondered why the stats package that generates these means and LSD values doesn't also group means automatically; I think I understand why now!

Hope this isn't driving you too crazy!

Thanks gain.

The fourth 'not working' example was bugging me after I'd posted it, your code got it right and I got it wrong manually: in my example, mean number 4 is in the same group as number 2 but it shouldn't be, your code is definitely right I checked and checked last night sorry!
Also, when I ran the second last example again I got a different result to the one in the spreadsheet- the pattern was spot on just muddled ranking like the other 2 examples-
I think you may have solved this for me and I went and confused you with wrong examples!

I do feel rather sheepish! sorry

Ho hum, you had actually convinced me that my method was not correct. I'll keep schtum from now on... See if this code is an improvement - it makes sure the groups proceed a,b,c etc rather than a,d,g.

Please Login or Register  to view this content.

You're very kind not to scold me!

Lots of complicated one's are working perfectly now, although there are a few more strange ones (attached again). the last macro worked better on these for some reason, I'm affraid I can't work out what the code's actually doing now so I couldn't pick out what the change may be doing.... I hope you can!

kind regards,

Graham

Thanks Stephen- progress is wonderful thank you, I'm very impressed and certainly grateful of all your time.
This latest code actually seems to just rank means a-g from smallest to largest, it doesn't look like it's using LSD correctly and so it's struggling with all the past examples that had previously worked.

What I was trying to do in those few examples that weren't working was change cetain aspects and see where the macro stumbled; it looks like the macro is more accurate when there are fewer decimal places involved.
I did wonder if examples with small differences, differences that were only just inside LSD, were throwing it off but the very last example in book 2 worked fine so I don't think that's the case.

Please don't feel limited by these examples; they are just made up numbers to fit an expected group order which I then run to see if what I expected comes back, if it helps to make up your own figures to incorperate into the model please do.

This has become a long week ofor you sorry- I may need to send you a new wax figure to replace the worn out one! and no hurry don't panic!

regards,
Graham

Graham - oh dear, back to the drawing board. I thought it was producing the right results, but maybe was unclear. What would be helpful if you could make absolutely clear what the results should be for those examples and then I can be sure I am working to the right end.

I've tried not to confuse things with 'what the macro got/ what it should have got' and just included examples as they should turn out, hopefully this helps.
I think it's almost there as a set can work perfectly, then if I add more units after the decimal on either the means or the LSD confusion arises.
So far I get the best results with the second edit of code if that helps?

you'll be seeing little a's and b's in your sleep all weekend!

Quite likely. I get most of my best ideas at 4am.