Given partial data about an exponential growth curve, calculates various parameters about the best fit ideal exponential growth curve.
Sample Usage
LOGEST(B2:B10,A2:A10)
LOGEST(B2:B10, A2:A10, TRUE, TRUE)
Syntax
LOGEST(known_data_y, [known_data_x], [b], [verbose])
-
known_data_y
- The array or range containing dependent (y) values that are already known, used to curve fit an ideal exponential growth curve.-
If
known_data_y
is a two-dimensional array or range,known_data_x
must have the same dimensions or be omitted. -
If
known_data_y
is a one-dimensional array or range,known_data_x
may represent multiple independent variables in a two-dimensional array or range. I.e. ifknown_data_y
is a single row, each row inknown_data_x
is interpreted as a separated independent value, and analogously ifknown_data_y
is a single column.
-
-
known_data_x
- [ OPTIONAL -{1,2,3,...}
with same length asknown_data_y
by default ] - The values of the independent variable(s) corresponding withknown_data_y
.- If
known_data_y
is a one-dimensional array or range,known_data_x
may represent multiple independent variables in a two-dimensional array or range. I.e. ifknown_data_y
is a single row, each row inknown_data_x
is interpreted as a separated independent value, and analogously ifknown_data_y
is a single column.
- If
-
b
- [ OPTIONAL -TRUE
by default ] - Given a general exponential form ofy = b*m^x
for a curve fit, calculatesb
ifTRUE
or forcesb
to be1
and only calculates them
values ifFALSE
. -
verbose
- [ OPTIONAL -FALSE
by default ] - A flag specifying whether to return additional regression statistics or only the calculated coefficient and exponents.-
If
verbose
isTRUE
, in addition to the set of exponents for each independent variable and the coefficientb
,LOGEST
returns-
The standard error for each exponent and the coefficient,
-
The coefficient of determination (between 0 and 1, where 1 indicates perfect correlation),
-
Standard error for the dependent variable values,
-
The F statistic, or F-observed value indicating whether the observed relationship between dependent and independent variables is random rather than exponential,
-
The degrees of freedom, useful in looking up F statistic values in a reference table to estimate a confidence level,
-
The regression sum of squares, and
-
The residual sum of squares.
-
-
Notes
- The statistics calculated by
LOGEST
are similar toLINEST
but use the linear modelln y = x1 ln m1 + ... + xn ln mn + ln b
for each independent variablex1 ... xn
. Therefore additional statistics such as the standard error must be compared to the natural logarithms of them
andb
values rather than the values themselves.
See Also
TREND
: Given partial data about a linear trend, fits an ideal linear trend using the least squares method and/or predicts further values.
LINEST
: Given partial data about a linear trend, calculates various parameters about the ideal linear trend using the least-squares method.
GROWTH
: Given partial data about an exponential growth trend, fits an ideal exponential growth trend and/or predicts further values.