wtf is the difference between precision and reliability (1 Viewer)

[anonymoushehe]

my teacher was going through terms for investigation skills, but im like so confused about these two terms

 

[liamkk112]

my teacher was going through terms for investigation skills, but im like so confused about these two terms

reliability = how reproducible and consistent is the experiment? usually boils down to two things:
- repetition: if the experiment is repeated many times and results are consistent, then clearly the experiment is reproducible and consistent
- control variables: the experiment should be repeated with the same conditions and method, and control variables should be the same through each repetition of the experiment - otherwise the results may be affected by external factors which were not accounting for, which could affect the experiment's reproducability and consistency (this is also linked to validity)

precision = how precise measurements are, this is basically affected by how good ur equipment is. eg if u use a ruler with 1cm increments to measure a length, ur measurements will be more precise than if u used a ruler with 10cm increments. usually this is measured by the uncertainty of a measuring device, which is basically how much of an error is possible in a measurement. for example, with a ruler with 1cm increments, we say the uncertainty is +/- 0.5cm. this is because we could be wrong by this amount in a measurement - if something was 1.5cm long, our ruler would tell us that it was 1 or 2 cm long if we go to the nearest increment, but in this case we were actually off by (+/-) 0.5cm. in general for any measurement in increments, the uncertainty is half the increment size (like with the 1cm ruler, there was +/- 0.5cm uncertainty)

precision is just the contributing factor to the overarching concept of accuracy which nesa loves, which describes how close your results are to the correct or accepted value. to improve accuracy, we can improve the precision of our measurements, the amount of significant figures we can measure, and also reducing systemic errors which can change our results by a small, but consistent amount, causing it to not be the accepted results/value.

id recommend getting very familiar with accuracy, reliability and validity, its very important in depth studies and nesa also loves to ask questions about it in the hsc in investigation questions. they are very annoying though

 

[anonymoushehe]

reliability = how reproducible and consistent is the experiment? usually boils down to two things:
- repetition: if the experiment is repeated many times and results are consistent, then clearly the experiment is reproducible and consistent
- control variables: the experiment should be repeated with the same conditions and method, and control variables should be the same through each repetition of the experiment - otherwise the results may be affected by external factors which were not accounting for, which could affect the experiment's reproducability and consistency (this is also linked to validity)

precision = how precise measurements are, this is basically affected by how good ur equipment is. eg if u use a ruler with 1cm increments to measure a length, ur measurements will be more precise than if u used a ruler with 10cm increments. usually this is measured by the uncertainty of a measuring device, which is basically how much of an error is possible in a measurement. for example, with a ruler with 1cm increments, we say the uncertainty is +/- 0.5cm. this is because we could be wrong by this amount in a measurement - if something was 1.5cm long, our ruler would tell us that it was 1 or 2 cm long if we go to the nearest increment, but in this case we were actually off by (+/-) 0.5cm. in general for any measurement in increments, the uncertainty is half the increment size (like with the 1cm ruler, there was +/- 0.5cm uncertainty)

precision is just the contributing factor to the overarching concept of accuracy which nesa loves, which describes how close your results are to the correct or accepted value. to improve accuracy, we can improve the precision of our measurements, the amount of significant figures we can measure, and also reducing systemic errors which can change our results by a small, but consistent amount, causing it to not be the accepted results/value.

id recommend getting very familiar with accuracy, reliability and validity, its very important in depth studies and nesa also loves to ask questions about it in the hsc in investigation questions. they are very annoying though

omg tysm i hate depth studies so much fml

 

[anonymoushehe]

wait im having a stroke again, my teacher is telling me that precision is how much the obtained results agree or align with each other but is this not literally reliability (consistency of results)??? I understand your explaination but now im like so confused again help

 

[ExtremelyBoredUser]

ngl im so proud of my depth study i did , brings back painful great memories.

 

[ExtremelyBoredUser]

wait im having a stroke again, my teacher is telling me that precision is how much the obtained results agree or align with each other but is this not literally reliability (consistency of results)??? I understand your explaination but now im like so confused again help

no.

accuracy = how much your value align with result
Reliability = consistency

they are not mutually exclusive. You can have a very accurate experiment with unreliable results. For the sake of understanding, this is not rigorous by any means and more just a way to understand the difference.

e.g
you do only one trial to measure gravitational acceleration.

first trial - 9.8 ms^-2

Now you can conclude that the g is 9.8 ms^-2 and this is indeed an accurate result since it aligns to the model value which is, say for this case, 9.81. But you could not conclude that this is a reliable result since its very possible that this might be a fluke.

in the second example, you do 5 trials

first trial - 4.9 ms^-2 (cbf writing units below)
second trial - 4.8
third trial - 4.85
fourth trial - 4.95
fifth trial - 4.98

Now this experiment yields reliable results since the difference between each value is relatively low. You might conclude g ~ 4.9 and this is reliable since such value is the mean of all the results. This value is clearly not accurate in this case since it significantly differs by the model value by ~50%.

in the last example, you do the 5 trials after realising you have an instrumental error

first trial - 9.8
second trial - 9.85
third trial - 9.75
fourth trial - 9.78
fifth trial - 9.82

The mean value in this case is 9.8 and not only is this a reliable result since the trial values are consistent with each other with minimal difference but also accurate since it aligns with the model value.


note:
generally 3-5 trials is considered as reliable for most physics experiments in high school (anecdotal) but ofc there's other stuff that makes results more reliable e.g line of best fit, ignoring outliers and so on but this is just to get you understanding why we differentiate reliability and precision in an experiment,

 

[moonsuyoung]

what everybody says above but if you want an image format ^^

 

[ExtremelyBoredUser]

View attachment 42411
what everybody says above but if you want an image format ^^

much better for visualising, i remember a similar image from one of the physics textbooks but cbf finding it.

 

[anonymoushehe]

no.

Precision = how much your value align with result
Reliability = consistency

they are not mutually exclusive. You can have a very precise experiment with unreliable results. For the sake of understanding, this is not rigorous by any means and more just a way to understand the difference.

e.g
you do only one trial to measure gravitational acceleration.

first trial - 9.8 ms^-2

Now you can conclude that the g is 9.8 ms^-2 and this is indeed a precise result since it aligns to the model value which is, say for this case, 9.81. But you could not conclude that this is a reliable result since its very possible that this might be a fluke.

in the second example, you do 5 trials

first trial - 4.9 ms^-2 (cbf writing units below)
second trial - 4.8
third trial - 4.85
fourth trial - 4.95
fifth trial - 4.98

Now this experiment yields reliable results since the difference between each value is relatively low. You might conclude g ~ 4.9 and this is reliable since such value is the mean of all the results. This value is clearly not precise in this case since it significantly differs by the model value by ~50%.

in the last example, you do the 5 trials after realising you have an instrumental error

first trial - 9.8
second trial - 9.85
third trial - 9.75
fourth trial - 9.78
fifth trial - 9.82

The mean value in this case is 9.8 and not only is this a reliable result since the trial values are consistent with each other with minimal difference but also precise since it aligns with the model value.


note:
generally 3-5 trials is considered as reliable for most physics experiments in high school (anecdotal) but ofc there's other stuff that makes results more reliable e.g line of best fit, ignoring outliers and so on but this is just to get you understanding why we differentiate reliability and precision in an experiment,

wait but i thought accuracy was how much the observed values align with the true value (in this case 9.81)?

 

[carrotsss]

wait im having a stroke again, my teacher is telling me that precision is how much the obtained results agree or align with each other but is this not literally reliability (consistency of results)??? I understand your explaination but now im like so confused again help

tbh if your teacher is the one marking it follow whatever ur teacher says

 

[nsw..wollongong]

pretty sure that
precision: how close your data points are to each other (eg. trial 1 = 9.00 cm, trial 2 = 9.10 cm, trial 3 = 9.05 cm)
reliability: getting the sm/similar values the more u repeat the test (eg. if someone else completed your tests, they would also get around 9cm)
accuracy: how close your data points are to true values (eg. ur average is 9.05 cm which aligns with Dr Seuss' reported 9.10 cm)

 

[ExtremelyBoredUser]

wait i

wait but i thought accuracy was how much the observed values align with the true value (in this case 9.81)?

Apologies, there was a typo, I copied paste some stuff from my notes.

Accuracy:
Measure of how close experimental values are to accepted true value

Reliability:
Extent to which repeated observations taken under identical trials yield similar results.

Validity:
Extent to which tests measure what was intended.

Note:
"Accuracy is how close a given set of measurements (observations or readings) are to their true value, while precision is how close the measurements are to each other."

Accuracy & Reliability were the big ones to note in my short answers.

 

[Luukas.2]

wait im having a stroke again, my teacher is telling me that precision is how much the obtained results agree or align with each other but is this not literally reliability (consistency of results)??? I understand your explaination but now im like so confused again help

Your teacher is wrong. Reliability is about consistency of results. Precision is about measurements, but also about reporting results. If you have data to 3 sig fig and you report your result to 6 sig fig, that is unjustified / false precision.

 

[wizzkids]

If you want to know exactly how NESA defines these terms validity, reliability and precision in scientific investigations, I recommend you download this document from the NESA website:
Evaluating Scientific Data Stage 4 - Stage 6