Logarithmic trans guide
Document Sample


Logarithmic Transformations for Exponential and Power Functions
Starting with a set of data, determine first which is the explanatory and which is response
Enter Explanatory into L1 and response into L2
Create a scatterplot with L1 as x and L2 as y
o Is it gently curved (could be power)
o Flat and then sharply curved (could be exponential)
Check the ratios between each response variable (divide by the prior and then divide by the increment
between x values
o If ratios overall are steadily increasing or decreasing, then power
o If ratios are approximately constant, then exponential
Exponential (x,logy) Power (logx,logy)
This is why it works: This is why it works:
(log both sides) (log both sides)
(apply log rules) (apply log rules)
(LSRL for (x,logy) (LSRL for (logx,logy))
(antilog) (antilog)
(exponential model) (antilog)
(power model)
This is how to do the process:
This is how to do the process:
Create L3 by placing cursor at the header and Create L3 by placing cursor at the header and
keying: log (L2) to fill in the values keying: log (L1) to fill in the values. Repeat to
Perform regression (Stat/Calc/8) for L1, L3, Y1 create L4 using log(L2)
write down the a and b values (to use later) Perform regression (Stat/Calc/8) for L3, L4, Y1
Check correlation coefficient (r) to see if strong Write down a and b values (to use later)
Make a scatterplot of L1, L3 with LSRL overlay to
Check correlation coefficient (r) to see if
see if regression worked strong
Test for linearity with a residual plot Make a scatterplot of L3,L4 with LSRL overlay
Use a = loga and b = logb to write the following to see if regression worked
equation: (step 3 Test for linearity with a residual plot
above)
Use a = loga and b = b to write the following
Perform antilog (back-transform) to create the
equation:
exponential model (steps 4 & 5 above)
Perform antilog (back-transform) to create the
exponential model (steps 4-6 above)
Get documents about "