Error analysis in log space: How to handle exponents in regression analysis?

[VINAYBAR]

Homework Statement

Maybe, I am a little stupid, but I just can't understand what I am doing wrong here. I have 2 quantities which I have to convert to log (base 10) and perform a regression analysis.

Quantity=1.235e9
error=3.4475e8

Homework Equations

I have to express the above quantity as log (base 10) with the error. The catch is I do not want to count the exponent i.e. just take log of the quantity and the error directly. I know, that if I take the log with the exponent then there is no problem, but I need to compare some results and hence don't want to do it.

The Attempt at a Solution

So, I took the log of 1.235=0.0917

The error in log space by the standard formula is 0.12. Now, if I count the powers then the quantity is 9.0917 and the error is unchanged. If I use my regression program with these values, then I get a decent fit. But, if I use it with a value of 0.0917 and error 0.12 then my fits are all messed up.

If something is unclear, please let me know and I can supply you with more information.

 

[n.karthick]

Why are you excluding the exponent? If you exclude then it will be a total different number
For example 1 can be written as 0.01e2
log of 1 is 0 and log of 0.01 is -2.
So both are not same. You should not avoid exponents while taking log because it is integral part of that number.

 

[VINAYBAR]

I have to ignore the exponent because after my regression is complete, I want to compare it to existing data. The previous data set does the calculation by ignoring the exponent i.e. (log (value))*(10^9) which is why I am ignoring the exponent as well.