Precision Vs. Accuracy

Some people are impressed by web sites which do a lot of very precise math on data with much lower accuracy. When they see a lot of math and technical jargon, they assume that the writer must be doing really good science.

I’ve been doing science and engineering professionally for the last 35 years, and take a different approach. I do sanity testing and if it doesn’t pass the sniff test, report the results of the sanity testing. It is malpractice to do very precise calculations on data produced by a methodology with very low accuracy.

Should I throw in a few random technical sounding words to impress people?

About these ads

About stevengoddard

Just having fun
This entry was posted in Uncategorized. Bookmark the permalink.

24 Responses to Precision Vs. Accuracy

  1. Robertv says:

    Please don’t because people like me would no longer understand what you are talking about if it is not about soccer.

  2. Beyond three significant figures you are pretty much guessing when using ordinary everyday data.

  3. bit chilly says:

    whenever i see numbers relating to the climate expressed to more than one decimal place i really do start to wonder,particularly relating to temperature.

    • ccglea says:

      You can never convince me that someone can calculate the worlds average temp to the nearest tenth, much less hundredth of a degree. When I see info that claims that precision, it fails my sanity check.

  4. gator69 says:

    “If you can’t explain it to a six year old, you don’t understand it yourself.”
    -Albert Einstein

    • _Jim says:

      I’ll remember that the next time one of the board’s greater intellects challenges the EM theory which underlies gas radiational physics … Oops, did I already exceed a six year old’s intellectual level?

  5. Anto says:

    I can show you a very impressive equation which PROVES that 2 = 1:

  6. ccglea says:

    I had a prof that used to chastise us for using more than one sig fig in our calculations. His reasoning? “The data’s no better than that!”.

  7. Leon says:

    After a rather minimal amount of explanation, the entire CAGW scare simply dissolves….it no longer passes the “sniff-test”.

    That is why politicians and their minions cannot allow sniff-testing when it comes to CAGW.

    That is why Gina McCarthy, in her victory laps regarding the latest executive action concerning coal power plants, used the following fabricated anticipated results of the executive order:

    1. It will prevent 100,000 asthma attacks per year (laughable).
    2. It will prevent 2100 heart attacks per year (a bold-faced lie).
    3. The increased electrical costs, which will far exceed her “official numbers”, are not a negative. THEY ARE AN OPPORTUNITY FOR (POOR PEOPLE) TO CONSERVE!
    4. That is why cannot she cannot mention the fact that other countries, notably China ad India, already have plans for the construction of 1200 new coal-powered electrical generating plants. These new coal-powered plants will dwarf any potential CO2 savings we may have from shutting ours down.
    5. That is why she cannot mention that some of the “Greenest” nations of the European Union, after their financially-disastrous policies towards “green energy” are harming their economies, are now turning back to coal as a source of energy for their electrical-generation plants…At precisely the time our President and McCarthy have decided to shut ours down.

    That is also why Gina McCarthy cannot mention the fact that the EPA’s own climate model, which demonstrably is exaggerating the effect, shows that these executive orders have a maximum potential of lowering the earth’s temperature by a whopping 18 one-thousands of one degree Celsius by the year 2100, while on the flip side they will do so much harm to our electrical generation infrastructure and economy that it will take a generation to reverse (if we can survive another generation with our current policies).

    A comment on the 100,000 asthma attacks: I will admit that am a sucker (and I think that is a good thing) when it comes to giving to charitable or other worthwhile causes.

    It is my impression that the American Lung Association, which has received millions and millions of dollars from the EPA, is the “Authoritative Source” behind the “100,000” asthma attacks” claim. It is also my understanding that this is not the first time that they have lent their name allowing politicians to “support their argument”.

    If someone can convince me that I am wrong about the ALA’s involvement in this issue, please do so now.

    Lacking that, financially supporting the people who are leading the fight against the CAGW cause is a much better use of your charitable dollars.

  8. dmmcmah says:

    “Climate Science” hides behind unnecessary complexity. That helps deflect criticism, makes them look smart, hide tampering, and satisfies their own inner need to be as smart as the theoretical physicists and electrical engineers. It also helps them convince themselves of their own lies.

  9. Chip Bennett says:

    The entire hockey stick model is based on this conflation of accuracy vs. precision. Essentially, the accuracy is minimized (“adjusting” current and historical raw data), which creates (unpublished) margins of error around the raw data; and then an unjustifiably precise trend is assumed through the error, in order to generate an exaggerated warming slope.

    That’s what happens when you don’t understand (or do understand, but intentionally abuse) precision vs. accuracy.

    • roger says:

      Unless it has been changed a WMO compliant thermometer has a basic accuracy of ±0.5º C. Remember Langmuir’s 2 and 3rd symptoms of Pathological Science:
      2.The effect is of a magnitude that remains close to the limit of detectability, or many measurements are necessary because of the low level of significance of the results.
      3.There are claims of great accuracy.

    • Leon says:

      The hockey stick also flew in the face of thousands of peer-reviewed studies that showed the medieval warm period was real, global and warmer than the present time. Nonetheless, when the hockey stick was presented to “2500 of the world’s top climate scientist”, it was immediately accepted as fact, while at the same time, the “scientists” who came up with this legend refused to release their computer code so the study could be replicated. That, in my opinion, does not even come close to passing the “sniff test”.

  10. Bob Greene says:

    Accuracy once was how wall the measurement matched the actual (true) value. Precision was how well the measurement reproduced the same or similar numbers. For example, a known 10g weight is placed on a balance and the weight is 10 gr, it is accurate. If the balance shows 9.5 grams, it is not accurate. If I weigh the weight 10 times and all the readings are nearly the same, the balance will be precise. If they vary widely, then the balance is not precise. Climate science seems to have different definitions.

  11. Steve Case says:

    Years ago I had a job where we tested the biotic activity of an industrial lubricant in terms of bacteria and fungi. The fungi was always way less than the bacteria by several orders of magnitude. Our instructions were to report a total biotic activity, bacteria plus fungi. The results often looked like this 2.0003 x 10³ per ml total biotic activity. I told the manager that the 0.003 tacked on to the 1 x 10³ made us look like idiots, but ignorance won the day. Think of Dilbert and you’ve got the idea.

  12. GaryW says:

    The original statement from Steve was about Accuracy versus Precision. This is a confusing issue for many people. The base concept is that averaging a bunch of collected data together should be useful. For the sociology folks in trying to figure out what people think or do, averaging the results of investigating many people should provide a more Accurate result to their research effort than simply asking one random person. Of course, that will depend upon the specific subject area of the inquiry and the people surveyed. Averaging the results of a survey of 24 people half of which are inner city community organizers and the other half rural Tea Party members is not likely to be of much use.

    For data gathered by instruments, we begin with a piece of information not available from human subjects: the absolute Accuracy and Precision of the instrument being used to make a measurement. These are to independent characteristics of an measurement instrument. Precision is an indication of how far apart the instruments calibration marks or individual measurement increments are. Accuracy is how far those individual marks or increments are from absolute true values. A humorous example of high Precision but poor Accuracy would be one of the gag rulers for fishermen that show a greatly exaggerated length of a fish. An example of the opposite might be a machinist measuring steel with millimeter marking but each mark Accurate to a fifth of a millimeter.

    For measuring a trend in local or global temperature, we use instrumentation and not human opinion. The problem we have been faced with is that, with the exception of satellite measurements and more recently the US Climate Reference Network measuring stations, the absolute accuracy of our atmospheric temperature data is inadequate for detecting trends less than a degree or two Celsius over a measurement period. However, we must understand that is not the fault of the historic measuring system as it was never design or intended to detect subtle climate shifts over time. It just happens to be the best data available.

    Where conflict about this historic data originates is from the attempts to improve the accuracy of past data. There are many different methods that have been attempted to develop this improvement. All of them, of course, are based upon a concept of assuming an predictable error amount and direction at a specific time and adjusting a temperature value to remove it. These involve things like changes in time of day when measurements were taken (TOBS), measurement site location changes, measurement site physical changes, and maintenance issues. For the most part, these specific potential sources of measurement errors sources are very poorly documented in site records. Interestingly, when attempts are made to apply corrections for these issues, those corrections are often greater than the resultant calculated long term temperature trend. That of course means that it possible that that trend is merely an artifact of the assumptions made for the temperature corrections.

    So where does Precision come into this? The question of precision comes up when temperature averages and trends are listed with a numeric Precision of tenths or hundredths of a Celsius degree. Seeing a printed temperature trend number of +0.503 degrees Celsius tends to indicate to most people that it is a very Accurate number. It may be a number derived from statistically processing hundreds of thousands of thermometer readings and the computer printout might show even more digits past the decimal point than shown above. However, the truth of the matter is that the starting point in the past is less Accurate than plus or minus a degree Celsius and the recent end point not much better so the range of possible slopes for a line drawn from one end to the other is plus or minus two degrees Celsius!

  13. Bob Greene is correct. I graded for Dr. Gilbert Ayers who authored “Quantitative Chemical Analysis” (second edition, 1968). He defines PRECISION on page 115 and ACCURACY on page 15. Mr. Greene agrees with Dr. Ayers.

    • GaryW says:

      This probably needs a little clarification. With instrumentation, where hysteresis, drift, and mechanical wear come into play (as well as a vast array of other possible problems), Precision as described in that context may not be adequate. The characteristic you are describing in instrumentation terms would be called REPEATABLY. While the difference may be semantic, it is those semantic differences that are the source of some misunderstandings in discussions about climate science. Your chemist may assume temperature trends and values are more reliable than an instrumentation technician would given the same description of the data collection.

      • Bob Greene says:

        What you are describing should be taken care of by instrument calibration. If you are worried about drift, etc, then your calibration frequency should address that issue. I’ll stand by the definitions of accuracy and precision.

    • Bob Greene says:

      About that time I was reading from the gospel according to Skoog & West. One of my graduate minors was analytical chemistry, but for those courses they seemed to think everyone understood precision, accuracy, drift and all those fun things that make analyses uncertain.
      One of the industrial QC (chemistry) labs that reported to me had a large number of measurements at 25°C. When we bought thermometers for lab use, we checked them against 2 NIST traceable thermometers and accepted anything within +/- 0.5°C. The failures were tagged with the error and donated to local schools. Failure rate was about 15%.

  14. Sparks says:

    “Anyone who doesn’t take truth seriously in small matters cannot be trusted in large ones either.” ~Albert Einstein

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s