3

Upon reading a peer reviewed article, I found this:

A Chi-square test shows that the demographic characteristics of the sample represent Toyota GB’s hybrid customers (data obtained from the New Car Buyers Survey Britain 2007, provided by Toyota GB), with 2.7% accuracy and 99% confidence.

I totally understand the concept of confidence interval , but I have not seen the accuracy part side by side with the confidence interval, can you explain what it means here specifically, or please point to me where I need to be reading?

gung - Reinstate Monica
  • 145,122
Immortal
  • 41

1 Answers1

1

Here's my opinion on this. If you think a little about the question, the intuitive answer is this: while confidence interval describes the range of mean values for a sample, accuracy of a statistics describes margin of error in measuring values. This relation is not to be confused with related, but different, one of accuracy and precision: http://en.wikipedia.org/wiki/Accuracy_and_precision. Actually, I believe that these relations represent exactly the same things, with the only essential difference being reversed terminology - what your question and much statistical materials refer to as confidence interval and accuracy, the above-mentioned Wikipedia article (and potentially other materials out there) refers to as accuracy and precision (see Figure below).

enter image description here

A detailed and IMHO very good explanation of a relation of and difference between confidence interval and accuracy can be found in the Chapter 20 of the second edition of Julian Simon's book "Resampling: The new statistics": http://www.resample.com/content/text/24-Chap-20.pdf. All chapters are available online here: http://www.resample.com/intro-text-online.

More confusing (to me), but nevertheless good and more statistically-focused explanation is this: http://www.bioconsulting.com/calculation_of_the_confidence_interval.htm.

Finally, another relevant discussion can be found here on Cross Validated site: Narrow confidence interval -- higher accuracy?.

It should be noted that often some papers, including statistical ones, use the term "accuracy" to denote a different meaning in a specific context, for example, to describe the ratio of true results of a medical diagnostic test to all results: http://www.lexjansen.com/nesug/nesug10/hl/hl07.pdf.

Community
  • 1
Aleksandr Blekh
  • 8,666
  • 1
    thank you reply . The graph is very nice in showing the difference between precision and accuracy. I Just wanted to know the meaning of the 2.7% here (in the question) as it is still confusing me , does it mean that we are accurate about the value of the mean of the sample relative to population by (100-2.7=97.3%), or it means that we are sure that the sample value might deviate from the population by 2.7%? (they might be the same...).... – Immortal Dec 13 '14 at 13:48
  • @Immortal: You're welcome. I think that 2.7% in your quote refers to measurement precision (as in the figure above), which is synonymous to the term accuracy in some other sources (such as the book I referenced). Feel free to upvote my answer, if you like it. – Aleksandr Blekh Dec 14 '14 at 01:58