Mistake in 2005 HSC? (3 Viewers)

[helper]

Correct, the BOS definition does not mention reliability as anything to do with how closely the data resembles a bell-curve.

Depends how you read it
"the degree to which repeated observations and/or measurements taken under identical circumstances will yield the same results."

The more samples you take the closer it comes to a bell curve. That is the closer it is to coming to the same results. It doesn't say the closer it comes to a value.

 

[pkc]

The more samples you take the closer it comes to a bell curve. That is the closer it is to coming to the same results. It doesn't say the closer it comes to a value.

True.
However if you take this interpretation it ignores the width of the spread itself.

A large number of samples closely fitting a very wide bell-curve would rank as more reliable than a smaller number of samples done in an experiment which produced far tighter clustered results.

This doesn't make much sense when you look at it from the bull's eye theory which talks about width of spread, or clustering around the average, as a measure of reliability.

 

[helper]

Every measurement has an inherent random error, which we can never remove but can reduce, by repeating we are approaching this limit which is the spread. We can reduce this by changing our instruments or method but not eliminate it. That is what is approaching.

The Bull's eye theory never shows every data point in the same point, rather clustered around a point.
It and the BOS definition are focused on students that do not have an understanding of statistics, so is targeted for the audience. As you increase your knowledge, the level of understanding of the model changes.

The same as atomic theory, in schools an adapted Bohr model is still normally used at is is understandable by students and explains what is needed. More up to date models are not used because most students would not understand them.

 

[NightShadow]

Yes its all interpretation.. to me the interpretation that reliability is where i can obtain more of the same value in the results is good enough [which is achievable by repetition] but obviously pkc doens't want to deal with it this way

ok

continuing with the standard deviation, you see from your data values..right or wrong... that the standard deviation even your case is small... not only small... but small and more or less the same in basically all the cases

i will now introduce - standard error
"quote -it has been generally agreed by statisticians that if the difference between the arithmetic mean of the various sets of data is more than twice the standard error, the results are significant and cannot be attributed to mere chance- unquote"

which means that if the standard error is kept at a minimum the experiment has a greater coefficient of significance, which to me means that the experimental results can be taken as reliable - the rsults are significant

and since standard error is more of less the same and the denominator of standard error increases as data valeus increase, the standard error decreases with increasing values = greater coefficiency of significance

this all rides on how you interpret the meaning of significance - to me significance means reliability

its also general understanding that reliability relates to the results not the actual experiment performed

 

[NightShadow]

basically... the accuracy of the measurement values = reliability


as proven by the standard error which enables with more data values , greater occurrences of the same data value popping up which is exactly what the BOS definition describes

I think pkc the whole thing is that accuracy is involved...but accuracy in BOS to me..relates to the measurements... not hte measurement values... accuracy of measurement values... relates to reliabilty... that is my understanding

 

[helper]

Accuracy of measurements yes but not accuracy of the investigation. I think this is why the BOS is careful in using validity and reliability of an investigation but not accuracy.

 

[pkc]

basically... the accuracy of the measurement values = reliability

If this is the case, then both answers A and B must be equally correct.

 

[helper]

No because you are not increasing the accuracy of the experiment, rather you are increasing accuracy of the data.
If the experiment has the wrong method the experiment will be no more accurate than if you do it once. So specifically you are checking the reliability of the experiment.
The question asked about the experiment not the data.

 

[pkc]

Yes its all interpretation.. to me the interpretation that reliability is where i can obtain more of the same value in the results is good enough [which is achievable by repetition] but obviously pkc doens't want to deal with it this way

Fair enough, thats one way to look at it

However, how can someone make this interpretation and at the same time subscribe to the bull's eye model of things - where the size spread of data spread determines the reliability?

If you are saying that the bull's eye model is wrong, or that reliability has nothing to do with the size of spread of the data, then ...I gotta say...you have a lot of textbook writers to argue with.

Also, how reasonable is it to expect a Yr12 student to come up with a different interpretation to what textbooks give them?

 

[helper]

The bull's eye model shows one aspect of reliability, that is you want a small SD. It does not talk about sample size.

Almost every new text book and most old resources on experimental design talk about increasing sample size to increase reliability.

 

[pkc]

The bull's eye model shows one aspect of reliability, that is you want a small SD. It does not talk about sample size.

Almost every new text book and most old resources on experimental design talk about increasing sample size to increase reliability.

Possibly we are meant to interpret the BOS definition in these 2 ways.

However, what happens if we decrease reliability by spreading the data out (use a "blunt" instrument) but at the same time increase the number of times we do the experiment. Have we increased the reliability? Somehow I think that interpretation leads to complicated contradictions.

Bob Emery, Principal, St Pius X High School, Adamstown, NSW.
and Dr Mark Butler, Gosford High School
(http://webs.mn.catholic.edu.au/physics/emery/accuracy.htm
23/1/06.)

define reliability this way:

"In terms of first hand investigations the Board seems to define reliability as repeatability or consistency. If an experiment is repeated many times it will give identical results if it is reliable."

This definition says there are 2 types of experiments - the type that tend to give identical results when you repeat them (narrow spread) , and those that don't (large spread). Nothing about more results=more reliability.

 

[helper]

Possibly we are meant to interpret the BOS definition in these 2 ways.

However, what happens if we decrease reliability by spreading the data out (use a "blunt" instrument) but at the same time increase the number of times we do the experiment. Have we increased the reliability? Somehow I think that interpretation leads to complicated contradictions.

That's right, which is why a question like that would never be multiple choice.

Repitition has clearly been told as a way of checking reliability to goverment school teachers by the support unit who are in consultation with BOS.

http://www.curriculumsupport.nsw.edu.au/science/index.cfm?u=3&i=113

 

[pkc]

That's right, which is why a question like that would never be multiple choice.

Repitition has clearly been told as a way of checking reliability to goverment school teachers by the support unit who are in consultation with BOS.

http://www.curriculumsupport.nsw.edu.au/science/index.cfm?u=3&i=113

Great info, thanks.

However a few hiccups

The definition they give:

"The term reliability refers to the consistency with which we can confirm the result (in this case the temperature change). "

Only agrees with the interpretion that reliability is the spread of data or "consistency" at which results can be produced from an experiment (presumably if it is done a large number of times.)

Now the box down the bottom is an interesting one and I can definitely see your point that it is implying that whether or not you have repeated the results.

All I can say its a leap of interpretation that :

a) contradicts its own definition in the passage above.
b) leads to contradictions (what happens when repetition is increased but repeatability of instrument is decreased) .
c) adds an interpretation that is not given in any BOS documents. (remembering that the DET only services about 60% of schools).
d) obviously hasn't been taken up by a number of teachers - and I dare say students- (previous public websites).

Thanks for the info, I can see why teachers and students would be led to give the Reliability answer by reading it.