Welcome Guest! The IOSH forums are a free resource to both members and non-members. Login or register to use them

Postings made by forum users are personal opinions. IOSH is not responsible for the content or accuracy of any of the information contained in forum postings. Please carefully consider any advice you receive.



Go to last post Go to first unread
#1 Posted : 23 June 2022 12:03:58(UTC)
Rank: New forum user

Explain principles of statistical and epidemiological analyses of data, including the use of the normal and poison distribution

What on earth do they want? Sorry too much NVQ meaning my brain is fried.

Please someone help me get my head round this

#2 Posted : 23 June 2022 12:25:44(UTC)
Rank: Super forum user

Start by reviewing what you know (or can find out) about the normal distribution (also known as the Gaussian distribution and as the bell curve) and the Poisson distribution (not generally known as the poison distribution).

These two distributions are used in different circumstances which gives you opportunity to compare their uses in statistical studies (analysing data sets) and epidemiological studies (analysing population health data sets).

Then think of different data sets relevant to your subject which would best be analysed using one or the other, and use these to illustrate the principles and the use of the distributions.

I cannot recommend too highly the classic book "How to Lie with Statistics", which doesn't only explain how to lie with statistics, but how to detect such lies and the principles of doing statistics properly (without ever getting very mathematical).

thanks 1 user thanked Kate for this useful post.
peter gotch on 23/06/2022(UTC)
peter gotch  
#3 Posted : 23 June 2022 16:23:53(UTC)
Rank: Super forum user
peter gotch

Hi Melliker

Good advice from Kate and good that the NVQ is including an understanding of statistics and the lies and misleading information that all OSH practitioners should be wary of.

Many things in life are subject to the "normal distribution" - in general 95% of some measure of failure etc will happen within roughly two "standard deviations" of the mean frequency - so 2.5% occurring more than 2.5% sooner and 2.5% later than two SDs from the mean.

In contrast some things are subject to skewed variations so we get the "poisson distribution" with a large head at one end and a long tail - looking a bit like a two dimensional picture of a fish ("poisson" is French for "fish").

But back to the normal distribution imagine going to the shop to buy a 100g bar of chocolate. It won't actually weigh exactly 100g and the dietician might want it to weigh less but the lawyers want you to get your money's worth - hence Weights and Meaures legislation tells you how much less than 100g is legally permissible.

In truth you won't probably notice if your bar weighs only 98.5g.

Now consider the number of accidents in construction in a fictional country:

Year 1 - 100

Year 2 - 94

Year 3 - 99 - year on year up 5, tut tut

Year 4 - 88 - year on year down no less than 11, well done?

Year 5 - 87

Overall this might look like a downward trend but it might be explained by e.g. a downturn in the sector, OR might simply be statistical blips - so we start looking at "95% confidence limits" - just as the opinion pollsters do.

So, next time the HSE publishes its Annual Statistics I can tell you with 95% confidence that there will either be coverage of the encouraging reduction in the headline number of fatal accidents or a seriously concerning increase.

But equally I can confidently predict that it is very unlikely that either change is significantly significant on grounds of health and safety standards alone. 

"There are lies, damned lies and statistics".

#4 Posted : 24 June 2022 07:41:20(UTC)
Rank: Super forum user

Peter, your account of the Poisson distribution is a bit off.

It's not to be relied on for an assignment.

thanks 1 user thanked Kate for this useful post.
peter gotch on 24/06/2022(UTC)
peter gotch  
#5 Posted : 24 June 2022 13:41:24(UTC)
Rank: Super forum user
peter gotch

Kate, simplistic perhaps but probably good enough for my A Level Maths (a VERY long time ago!) but for most things in an NVQ assignment on health and safety (rather than applied mathematics or advanced epidemiology) I guess that applying the principles of a normal distribution would generally be close enough.

Edited by user 24 June 2022 13:42:20(UTC)  | Reason: Clarification

Brian Hagyard  
#6 Posted : 24 June 2022 14:17:29(UTC)
Rank: Super forum user
Brian Hagyard

Not sure if this helps but the dictionay definition says.  By definition, epidemiology is the study (scientific, systematic, and data-driven) of the distribution (frequency, pattern) and determinants (causes, risk factors) of health-related states and events (not just diseases) in specified populations (neighborhood, school, city, state, country, global).

The thing i was always told to be wary of was not linking 2 unrelated facts - so Edwin Chadwicks epidemiological study of Cholera in london linking it to specifc wells (even though they did not know the exact cause) is a classic example good research - while linking Autism to MRM vaccination is a bad example as no true relationship could be made - would have just as easy to say it was linked to any other factor - increased use of microwave ovans for eample.

Users browsing this topic
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.