Am I WRONG or is "STATISTICS WRONG"?!

#1
(Am I WRONG or is "STATISTICS WRONG"?)
YATES ANALYSIS (The “usual calculations” SEEM TO BE WRONG!)
These values could be considered as “idealised results” for 3 variables, each at 2 different levels.
For each “result” the variable levels for that particular result is [N;C value;B value;A value]
0 = low; 1 = high
1000, 1001, 1010, 1011, 1100, 1101, 1110, 1111
1001-1000 = effect of changing “A value” from high to low. Etc.
ALSO (for 2 way interaction)
1011-1000 = effect of changing “for 2 way interaction of AB value” from high to low. Etc.

STANDARD YATES ANALYSIS
The MAIN "difference" is measured from the "baseline"
The TWO WAY INTERACTION "difference" IS NOT measured from the "baseline".
(It seems to be measuring a "difference" between 2 "different TWO WAY INTERACTION results values" that have the same "levels" for the variables.)
A "variance ratio" should be comparing "very similar values". IT DOESN'T!

“usual calculations” (which give the same results as for ANOVA)
- give differences, from the “baseline reference value”, for the “MAIN effects”
- give differences, as “self-reference values”, between 2 of “the same interaction results”, for the “INTERACTION effects”.
“Variance ratios”, must, by their very nature (& mathematically) must have considerable similarity. For this"variance ratio", it’s like dividing the number of peas by the length of carrots! Meaningless mathematically! You can’t divide the number of teeth by the “number blue”, to get a ratio!
The “usual calculations”, below, has calculated VERY DIFFERENT (i.e. WRONG!) values for the “Yates Calculation Value” and hence the calculated “Yates effect”.

1cba-------------Yates--calc---------effect
1000---2001---4022---8444---8---1055.5
1001---2021---4422--------4---4--------1
1010---2201--------2-------40---4------10
1011---2221--------2---------0---2--------0
1100--------1--------0------400---4-----100
1101--------1-------20---------0---2--------0
1110--------1---------0---------0---2--------0
1111--------1---------0---------0----1-------0
##################
MODIFIED YATES ANALYSIS
The “MODIFIED calculations”, below, has calculated the values LOGICALLY (i.e. RIGHT!) values for the “Yates Calculation Value” and hence the calculated CORRECT “Yates effect”
The MAIN "difference" is measured from the "baseline"
The TWO WAY INTERACTION "difference" IS measured from the "baseline"
The "variance ratio" can be compared as they are "very similar values" of "the same type".
.(The "variance ratio" for the TWO WAY INTERACTION & the THREE WAY INTERACTION, will be "different" for these calculations, compared to the "usual calculations"!)

1cba-------------Yates--calc---------effect
1000---2001---4022---8444---8---1055.5
1001---2021---4422--------4---4--------1
1010---2201--------2------40---4-------10
1011---2221--------2------22---2--------11
1100--------1------20-----400---4------100
1101--------1------20-----202---2------101
1110--------1--------0-----220---2------110
1111--------1--------0------111---1------111

The "Modification factors":-
[row heading] [calculation for “Yates Calculation Value” ] [“Yates Calculation Value” ]
[1011] # [(1111-1000)+(1100-1011)] # [22]
[1101] # [(1111-1000)+(1101-1010)] # [202]
[1110] # [(1111-1000)+(1110-1001)] # [220]
[1111] # [1111-1000 ] ############ [111]
(This problem of "Yates effects" & "variance ratios" being "illogical", was found when "my simple statistical analysis" {independently derived from first principles} was DIFFERENT from the STANDARD YATES ANALYSIS! The values calculated from MODIFIED YATES ANALYSIS matches "my simple statistical analysis" values! "My simple statistical analysis" (fairly) easily calculates 4 "virtual results" for each of the 8 "experiments", allowing "in-depth statistical analysis"!)
[Formatting is a problem. The formatting seen during EDITING is different to what is seen after SAVE & both are different what is from a "wordprocessor COPY"!]
 
Last edited:

Dason

Ambassador to the humans
#2
No offense intended but is English your native language? Either way your writing style is not good. You aren't doing a good job of conveying the information you have and/or the questions you're asking. Please give rewriting this a shot and try to sound less crazy when you do.
 
#3
No offense intended but is English your native language? Either way your writing style is not good. You aren't doing a good job of conveying the information you have and/or the questions you're asking. Please give rewriting this a shot and try to sound less crazy when you do.
Please give rewriting this a shot - I hope the "reformatting" of the above, makes for a major increase in clarity!
Try to sound less crazy when you do - "Reinventing" "analysis of variance" sounds "crazy", but I had "a major insight" for increasing simplicity and a more "in-depth statistical analysis", at the same time!
 
#5
This is ""my simple statistical analysis". [It's a lot simpler when the "required for formatting 'help' " & the "unnecessary cells" are removed!) The formatting has now been improved! :)

These values could be considered as “idealised results” for 3 variables, each at 2 different levels.
For each “result” the variable levels for that particular result is [N; C value; B value; A value]
0 = low; 1 = high
1000, 1001, 1010, 1011, 1100, 1101, 1110, 1111

The cell values, “within the table” = [“row heading”] – [“column heading)
====1000==1001==1010==1011==1100==1101==1110
1000===0====-1===-10===-11==-100==-101===-110
1001===1=====0===-9====-10==-99==-100===-109
1010==10====11====0====-1==-90====91==-100
1011==11====10====1=====0==-89===-90===-99
1100=100===101==110===111=== 0====-1==-10
1101=101===100===111===110===1====0===-9
1110=110===111===100===101===10==11===0
1111=111===110===101====100===11==10====1

Only the “lower left”, “triangle” of results are needed!
11 [BLACK] is a VALID “virtual result”.
11 [RED] Has been made A VALID result, by , a modification to the value, at the next advanced level to create a VALID “virtual result”.(This then gives 4 VALID “virtual results” for EVERY Main effect; gives 4 VALID “virtual results” for EVERY two-way interaction; gives 4 VALID “virtual results” for the three-way interaction. Meaningful variances & variance ratios can be calculated from this.)
11 [RED] UNMODIFIED Is a NOT A VALID result, because the calculation (on the "binary type values") does not calculate a "low" {0} or "high" {1} value for a variable.
For “idealised results”, the VALID “virtual results”, along the “top left to bottom right”, diagonal are identical!
4 “virtual results” for each of the MAIN effects, from simple subtraction.
For “idealised results”, the VALID “virtual results”, along the “top left to bottom right”, diagonal are identical!
2 “virtual results” for each of the 2-WAY INTERACTION effects, from simple subtraction.
A "quick comparative look" at the variations in the numbers, gives a "thumbnail view" of what the results show.
If the “binary type results” are used in the above table, all the VALID results are also “binary type results”.
A“binary type result” show that the “virtual result”calculated is a “low” or “high” value for the variables
The NON VALID results are NOT “binary type results”

[Formatting is a problem. The formatting seen during EDITING is different to what is seen after SAVE & both are different what is from a "word processor COPY"!]
 
Last edited: