What's the difference between accuracy and precision in measurements or data? Why is it important?

This article defines the terms accuracy and precision as they are used to describe measurements of all kinds.

Whether you are measuring electrical resistance, the comparative depth of stair treads, or the number of airborne mold spores per cubic meter of air, reporting the results without including an indication of the margin of error of the measurement can make very precise numbers dangerously misleading.

What is the Difference Between Precision and Accuracy in Measurements

- Daniel Friedman

Watch out: don't confuse measurement precision with measurement accuracy. In the expression of measurements, precision refers to the number of decimal places or digits in a number obtained by the measurement, while accuracy describes the margin of error in the measurement.

Article Contents

• ACCURACY vs PRECISION of MEASUREMENTS
  • Typical Sources of Errors Found in Most Measurements
  • Errors when using a measuring tape - measure twice, cut once
  • Errors When Measuring Electrical Data: Volts, Ohms, Amps
  • Errors When Measuring or Counting Airborne Particle or Mold Spore Levels
• CONVERT PPM to % CONCENTRATION - separate article

 

People who do not understand this precision - accuracy distinction can be misled with regard to the reliability (accuracy) of numbers that are presented with much precision if the margin of error in the measurement was significant..

129.4 is a number that is less precise than 129.43939480

But if the possible range of error in our measurement is 10%, then our measurement of 129.4 OR our measurement of 120.2939480 both could be expressed as +/- 12 (since 12 is 10% of 120). This means that the range of accuracy of our measurement of 129.43939480 +/- 10% means that

at a 10% range of error (or for 129.4, +12.9V / - 12.9V)

the actual or "true" number could be anywhere between 133.43939480 and 108.43939480

which makes those extra decimal points meaningless, perhaps even misleading

To avoid presenting misleading results about the accuracy of our measurement, in most circumstances we would not report the measurement with all those decimal places. We would report 129.4 +/- 10%.

Typical Sources of Errors in All Measurements

1. Device accuracy: One source of error in making any measurement is the accuracy of the measuring device itself.
2. Procedural effects: A second source of error in making any measurement or really a collection of possible errors is found in the operation or use of the measuring device - human-induced errors or human variables (not all measurement errors are fairly called "mistakes")
3. Environmental effects: A third source of error in making any measurement are environmental or test condition variables such as temperature or air movement variations that affect the release of spores from a mold reservoir in a building into the building air

Errors when using a measuring tape or ruler

When using a typical carpenter's or engineer's measuring tape, the end tab on the measuring tape is left deliberately loose (by the manufacturer) in order to improve measurement accuracy by allowing for the thickness of the tab itself when making an inside measurement compared with making an outside measurement.

At left our photo shows some of the author's measuring tools used in general carpentry. The accuracy of measurements made with each of these devices depends on how carefully the device is used, placed, and read.

If you are measuring the length of a board a good approach is to hook the measuring tape tab over the end of the board at the starting corner, run the tape parallel to and along the same edge of the board as the starting corner, and read the measurement by looking straight down at the tape from the other end of the board.

Such measurements made with a measuring tape are easily accurate to 1/16" - that is the measured length is typically correct +/- 1/16" in the hands of most carpenters. The potential errors that this approach must address include:

• Misplacement of the measuring tape's hook-tab at the starting end of the board. If a helper holds the starting point or "dumb end" of the tape with the tab pushed towards the carpenter who is pulling the "smart" end from which the measurement will be made, this introduces a small error of about 1/16" - not significant for wood frame construction but potentially important in other applications.
• Misalignment of the measuring tape. A beginner who hooks the measuring tape at one corner of the starting end of a board and pulls the tape on a diagonal across to the other end of the board to obtain a length measurement is actually measuring the board's diagonal - introducing a length-error measurement that is more significant (as % error) over wider, shorter board measurements.
• Mis-reading of the measuring tape. Measure twice, cut once, is a good adage especially when working with valuable lumber or when you are cutting the last piece with no extra material to cover up an embarrassing error. Working in a rush it's easy to make significant errors in reading the tape - by looking at an angle rather than straight down at the tape marking over the board edge or end, or by working too fast and simply mis-reading inches or feet.

Errors When Measuring Electrical Data: Volts, Ohms, or Amps

When using an electrical measuring instrument such as a DMM to measure amps, volts, or ohms, the instrument display window will indicate some number of decimal places in the result (precision) while the instrument's specifications will tell you the accuracy (typically +/- 1%) for the instrument function and varying depending on range settings.

When measuring the voltage level of an electrical circuit in our office in Mexico we find that from time to time the actual voltage level can vary by about 10%.

So while we might report that at a given measurement we measured the voltage level at 108V, that measurement should be presented as 108VAC +/- 10%, telling our client that over time voltage in that location typically varies between 98VAC and 118VAC (excluding periods of power loss when voltage is zero).

At left our photo illustrates the Sperry DSA-500 clamp-on multimeter/ammeter reading 0.28A (Amps) of current flow on an electrical circuit. The function dial has been set to the 0-40A range, the proper ampacity testing range in order to report with finer precision when measuring lower current levels than if we had used the 400A range setting. But what about accuracy differences in different DMM measuring ranges?

Sperry's documentation for this DMM indicates that the accuracy of the instrument when measuring AC current (Amps) is +/- 2.0% rdg+/-6dgt when making measurements at 23+/-5degC and 45-75% relative humidity. The company is in essence warning that at more extreme temperatures or humidity levels the accuracy of the instrument may vary from this level.

At DMM accuracy, typical specifications we provide a table describing typical DMM accuracy specifications for its different functions.

Details of the procedure for measuring amps are provided below at How to make current measurements.

Also see DMM DIGITAL MULTIMETER HOW TO USE.

Errors When Measuring or Counting Airborne Particle or Mold Spore Levels

Airborne Particle Sampling Procedure Errors or Inconsistencies

In addition to this catalog of sources of errors in airborne particle or mold spore test results, for more in-depth information on tihs topic please see ACCURACY OF AIR TESTS for MOLD and the technical articles listed there.

When measuring the concentration of airborne mold spores inside a building space using a spore trap, the laboratory reporting the sample analysis results may report that 1034 spores of Penicillium/Aspergillus mold were detected.

But our own careful study of spore traps and air sampling equipment and procedures found that the accuracy of that measurement can vary by one to three orders of magnitude depending on exactly how and when the sample was collected.

Effects of Site Activity During Airborne Particle or Mold Sampling

For example, walking across a carpeted floor during sampling can double or triple the airborne particle level, and turning on an overhead fan or an HVAC system can increase the airborne particle level by three orders of magnitude.

That means that our mold air test report result of 1034 spores of Penicillium/Aspergillus mold per cubic meter of air is a description of a mold level that in actual experience in the building could be as high as 100 times that number, or 103,400 spores of Penicillium/Aspergillus mold per cubic meter.

If the true airborne mold exposure level in the building is reaching that higher level the meaning of the mold test is completely reversed from "no problem detected" to "severe mold contamination detected".

There is normally high variation in the level of airborne particles from moment to moment in buildings. Actual field data easily demonstrates that particle presence in indoor air varies by orders of magnitude from minute to minute as our Allergenco impaction air sampler slide illustrates (photo at left).

Sampling identical quantities of air indoors just a few minutes apart regularly shows up in our data as enormous differences in particle density from interval to interval, as you can see in this photograph of parallel traces of airborne particles captured by an air sampler which collected these samples just minutes apart in the same location in a building.

We always see this phenomenon in buildings, since unless we are measuring airborne particles released at a fixed rate in a controlled test chamber there are quite a few site conditions that agitate airborne debris.

For details see Extent of Variation of Airborne Particle Counts. Also see AIRBORNE MOLD SPORES CONCENTRATION BURSTS.

Effects of Type of Sampling Device Type, Sample Duration, Actual Particle Levels on Airborne Particle Level Detection

The most common airborne particle sampling devices in wide use for indoor air monitoring (such as the detection of airborne mold spores) rely on an inertial impaction sampling cassette or "spore trap" connected to a calibrated air pump that moves a specified volume of air through the sampling device over a short interval, typically just a few minutes.

These slit impactors rely on a venturi through which airborne particles accelerate and then impact and stick to a small slide that contains a sticky collection medium.

We have already explained that enormous variations in site conditions can produce orders of magnitude variation in the actual level of airborne particles there to be "detected".

But in addition, design presumptions and site conditions can have a significant effect on the ability of of these devices in the field to accurately collect airborne particles.

For example, if the level of airborne particles is very high and the measurement or collection interval too long, the collection slide becomes overloaded with particles.

Not only is such a collection impossible to read accurately in the microscope, worse, no one will have the slightest idea how many particles arrived so late in the sampling interval that they just bounced off of the overloaded slide rather than sticking to it - as earlier particles have covered the sticky collection medium.

At AIRBORNE PARTICLE TEST SAMPLING CASSETTE STUDY we describe the use of typical spore trap sampling devices for indoor air measurements (photos above and at left).

Where greater accuracy is required we must therefore shorten the sampling interval enough that we are sure that we have not overloaded the collection medium - a decision that can be made subjectively in the field if the investigator is experienced in both field and lab procedures and if s/he takes the time and trouble to make an initial visual examination of the collection device at the end of the sample.

We suspect that among those mold test consultants who simply collect samples and send them to a lab (without an accompanying detailed site investigation) ever exercise such judgment.

Using a filter type collection device that captures 100% of the particles (in a given size range) during the sampling period can avoid the particle loss problem, though analysis of such samples may be more difficult, time consuming, and costly.

However we rarely find field reports of building investigations that make use of this latter method.

Effects of Sampling Methology In A Building on detected levels of airborne particles

Air sampler placement: Our photos below illustrate large differences in the number of airborne Aspergillus sp. particles collected at an indoor location. This variation occured when the only parameter we changed was the height of the sampler above the floor. The mold reservoir was on the under-side of a pool table at the sampling location. At below left is the particle density on the sample collected on top of the pool table, and at below right is the particle density on the sample collected on the floor below the pool table. Other test conditions were kept the same.

Details are at Particle Levels vs Sampler Height.

Opening a window or rapping on a heating duct: Below we illustrate a similar effect - a change in airborne particle levels caused by drafts ensuing after opening one window in a high-rise office building. Details are at Particle Levels vs Windows/Doors. Similar large changes can be observed when testing HVAC systems depending on whether or not the ductwork is disturbed. For full details of this topic see AIRBORNE PARTICLE & MOLD LEVELS in DUCTWORK.

Effects of Sampling Interval: use of long term or continuous working level monitors for airborne particle levels

One approach to smoothing the variability in measurements of airborne particle levels over time is the use of continuous working level monitors, recording instruments, or for monitoring outdoor pollen, mold, and other particle levels, use of special sampling devices (Anderson cascade impactor, quartz crystal cascade impactor, or filtration type particle collectors) that collects airborne particles continuously, depositing them on a collection tape surface mounted on a revolving drum.

These approaches increase the short term accuracy of airborne particle studies but unless data is collected over a much longer interval than 24 hours, the results will not reflect important variations that occur during changes in season and weather (temperature, humidity, wind direction, etc.) as well as local events such as traffic and exhaust levels or trash / agricultural burning.[6]

Effects of Laboratory Procedures on Airborne Particle or Mold Counts

The photograph shows four airborne mold trace samples collected inside the same hot air heating furnace plenum. We varied conditions from passive to more aggressive sampling methods during the brief one minute sampling interval.

Even before microscopic examination or counting the spores per cubic meter of air, it is visually obvious that there is very wide variation of particle collection level among these samples.

Laboratory "particle count" procedures may also vary among test labs, though ours and many other forensic labs regularly participate in "round robin" comparative sample analysis and other procedures that improve laboratory reporting procedures and accuracy.

See AIRBORNE PARTICLE ANALYSIS METHODS.

Field and laboratory data recording and even the most basic examination of the results demonstrate that very significant variability in field conditions and in particle behavior make airborne particle counts extremely inaccurate except under controlled conditions such as in a research chamber.

Similar technical shortcomings raise serious questions about the use of mold cultures, whether by settlement plate, swab, or Andersen sampler, to characterize an indoor mold level.

Absence of Reporting of Field Measurement Conditions Confounds the Meaning of IAQ Tests

Finally, the almost total absence of recording of site conditions at the time of such measurements adds another almost overwhelming degree of inaccuracy to this approach to characterizing mold exposure risk in buildings.

For example, simply turning on a fan or walking through a room during an air sampling procedure can completely change the results of the measurement. While there is a useful place for these tools, their application as simple tools to make a statement regarding mold exposure levels in buildings appears to be highly questionable.

Thus airborne mold exposure levels based on single-time-interval use of these tools are unlikely to be accurate.

Forensic Laboratory or Aerobiology Laboratory Airborne Particle or Mold Spore Count Procedural Inaccuracies

Additional inaccuracy in airborne particle counts or mold spore levels per cubic meter of air produced by a test lab can occur depending on both lab procedures and by normal and quite understandable human error.

Lab procedure variables include the percentage of spore trap particle trace that is counted and whether or not the lab analyst examines the entire trace at lower magnification to identify inconsistencies before beginning counting.

Add procedure variation in how, for partial-trace-counts, the analyst moves the microscope slide on the stage, how the microscope has been calibrated and adjusted, and how the lab sample has been prepared and mounted - all factors that can affect the visibility of individual particles.

More confusing, some analysts may count a cluster of spores as "one" particle while others (as we recommend) count or estimate the total number of particles in the cluster.

 

...

Continue reading at ACCURACY OF AIR TESTS for MOLD - topic home, or select a topic from the closely-related articles below, or see the complete ARTICLE INDEX.

Or see these

Recommended Articles

• ACCURACY OF AIR TESTS for MOLD - home
• ACCURACY vs PRECISION of MEASUREMENTS
• AIRBORNE PARTICLE TEST SAMPLING CASSETTE STUDY
• AIRBORNE MOLD COUNT VALIDITY
• AIRBORNE MOLD SPORES CONCENTRATION BURSTS
• CULTURE PLATE MOLD TEST ERRORS
• DRYWALL MOLD TESTING
• FALSE NEGATIVE MOLD TEST RESULTS
• FORENSIC LAB TECHNICAL PROCEDURES
• INDOOR vs OUTDOOR AIRBORNE MOLD COUNTS
• MOLD LEVEL IN AIR, VALIDITY
• MOLD TESTING METHOD VALIDITY
• MSDS Sheets for Gulf Oil Spill Dispersant Corexit - sampling errors
• AIRBORNE PARTICLE & MOLD LEVELS in DUCTWORK
• VALIDITY of MOLD TESTING METHODS
• VARIATION IN AIRBORNE PARTICLE LEVELS - CAUSES
• VARIATION IN AIRBORNE PARTICLE COUNTS vs SAMPLER HEIGHT
• VARIATION IN AIRBORNE PARTICLE LEVEL-COUNTS, EXTENT-OF

Suggested citation for this web page

ACCURACY vs PRECISION of MEASUREMENTS at InspectApedia.com - online encyclopedia of building & environmental inspection, testing, diagnosis, repair, & problem prevention advice.

Or see this

INDEX to RELATED ARTICLES: ARTICLE INDEX to BUILDING & HOME INSPECTION

Or use the SEARCH BOX found below to Ask a Question or Search InspectApedia

Ask a Question or Search InspectApedia

Try the search box just below, or if you prefer, post a question or comment in the Comments box below and we will respond promptly.

Note: appearance of your Comment below may be delayed: if your comment contains an image, photograph, web link, or text that looks to the software as if it might be a web link, your posting will appear after it has been approved by a moderator. Apologies for the delay.

Only one image can be added per comment but you can post as many comments, and therefore images, as you like.
You will not receive a notification when a response to your question has been posted.
Please bookmark this page to make it easy for you to check back for our response.
Our Comment Box is provided by Countable Web Productions countable.ca

Citations & References

In addition to any citations in the article above, a full list is available on request.

---

ADVERTISEMENT

---