LINEST further remarks

[Regina Henschel <rb.henschel@t-online.de>]

Hello members,

in addition to my previous mail about missing multiple regression [1], I 
will point out same additional problems:

(1)
<cite>
knownY: The set of y-values for the equation, either one single column 
or one single row
</cite>
Excel, Gnumeric, and OpenOffice.org allow more than one column or row in 
knownY, if the corresponding array knownX has the same size. Under that 
condition, they perform a simple linear regression as if the knownY and 
knownX values would be arranges in one column or row.

(2)
<cite>
knownX: ... If one single value, it is applied to all.
</cite>
I do not understand it. Does it mean, that this single value is used for 
each value of knownY? Then all y-values would result from the same 
x-value and there would not exist a best-fit linear function.


And now a lot of editorial remarks:

(3)
In table 30 it should not be r² but R² because it is the "coeffizient of 
determination" which is not the squared "sample correlation coefficient" 
in case of multiple linear regression.

(4)
In the equation after table 30 the index should not be n but k, because 
above n is used for COUNT(knownY) and because k is used later on in the 
definition of degrees of freedom.

Same confusion in equation for degrees of freedom and for F-Statistic. 
Use always n=COUNT(knownY) and k=number of variables of the regression.

Do not mix n and N, but use always lower case or always upper case.

(5)
Standard Error of the Slope (a) value is calculated using this formula
has to be:
Standard Error of the Intercept (a) value is calculated using this formula

Standard Error of the Intercept (b) value is calculated using this formula
has to be:
Standard Error of the Slope (b) value is calculated using this formula

Both occur two times.

(6)
Use consistent terms for "hat y_i".
In formula of line "The RSQ Value (r2) is calculated using the following 
formula" the equation "y_calc=a+bx" is used, but in formulas of line 
"The Regression Sum of squares is calculated as follows" and of the line 
"The Residual sum of squares is calculated using this formula" the 
equation "hat y_i=a+bx" is used.
In Addition both lack the index of x.

In both cases it should be "hat y_i = a + b x_i"

(7)
In the text part for the case "const = FALSE()" the definition of the 
Degrees of Freedom should be placed before the definition of "Standard 
Error of the Y value", because it is used there.

(8)
In case "const = FALSE()" the formula for slope b lacks indices for x 
and for y.

(9)
In case "const = FALSE()" in the formula for F-Statistic use k instead 
of N-df for to be consistent with the case "const = TRUE()".

kind regards
Regina

[1] http://lists.oasis-open.org/archives/office-comment/201006/msg00007.html