Wednesday, September 06, 2006

Decimal Points and Accuracy...

...are not exactly equivalent.

This is a common misconception. Many clients demand more decimal points to the reported data saying it'll be 'more accurate'. No! This is not an accurate statement to make (pun intended, thank you)

A radio announcer saying that "Today's temperature is going to be 70.83018460288453 degrees F" does not make that estimate accurate!

That 70.83018460288453 may be very far from the truth of say, 60 deg. F, though the measurement is precise.

Here's a quote from 'Common Errors in Statistics' , Good and Hardin, 2003 (Glossary on page 187 ) :
"An accurate estimate is close to the estimated quantity. A precise interval estimate is the narrow one. Precise measurements made with a dozen or more decimal places may still not be accurate"

8 comments:

Suspect said...

Nice one. Subtle distinction.

>Many clients demand more decimal points to the reported data saying it'll be 'more accurate'.

Sounds like the pointy haired boss. Fool them!

What about accuracy as it pertains to measurement (versus estimation)? Excessive preciseness may actually be harmful, because someone who sees the measurement might take the accuracy to be more than what it actually is.

Which is why we have those lessons on significant figures and such first thing in physics.

Dr. Katte's Blog said...

Good clarification about accuracy and precision! Our students calculate and express the final result using 9 decimal palces, even when given data have only 2 or 3 decimal palces!
Many engineers conducting experimental investigation are unaware of these, and also uncertainty analysis. I have a post on this
http://sskatte.blogspot.com/2006/08/course-in-measurements-in-mold-of.html

Sharan Sharma said...

Hi Kshitij,
> because someone who sees the measurement might take the accuracy to be more than what it actually is.

You meant Precision? :)

Hi SSK,
Saw your post, and liked the link to history. Yes, if everything was only about measuring things right...:)

Ajith said...

the other loosely used term is validity v/s reliability. Its somewhat related to what you've written I guess....most people say we have a reliable sample what they intend actually is we have a valid one!

Arunn said...

An accurate post clarifying accuracy from precision...;)

Sharan Sharma said...

Thanks Arunn.
Actually one more post on this coming up. Hope that makes it clearer.

Trundling Grunt said...

Too many people focus on the nth decimal place for data that can barely be resolved at the whole number level. It's an illusion but one people feel comfortable with it.

Sharan Sharma said...

Very true TG! Thanks for dropping by.