Statistical treatment of occupational heat and vibration assessments

Hello everyone.

I have a question about the statistical treatment of vibration and heat measurement results, as here in Brazil there is somewhat resistance to the use of AIHA’s IHSTAT spreadsheet for this type of approach, and the use of this tool is more common in quantitative assessments of noise and chemical agents.

The justification for not using the AIHA IHSTAT method is because the vibration and heat data are obtained in a biased way, that is, obtained at previously known moments and therefore the data could not be represented in a random way, as are the doses. of noise and the time-weighted average concentration of chemical agents.

Can you give your opinion on this? And, if possible, provide me with another approach to these two aforementioned physical agents.

Best regards.

Gustavo Rezende
Occupational Higienist

Hello Gustavo,

I have no experience whatsoever with these types of measurements :slight_smile:

That acknowledgement being out of the way, here are some thoughts:

If one measures a phenomenon ( chemical concentration in air, noise, vibration, heat, UV exposure) with the intention of applying their observations to moments not the object of measurements (e.g. measuring UV exposure during a couple of days and making a decision on long term risk), there is some inference going on.

This inference can be entirely subjective : I believe these measurements mean that…and this is my opinion based on my thought process

This inference can also be statistical : based on statistical theory, these measurements suggest that…In that case you need a target population from which your measurements represent a reasonably independant and random sample.

If you decide to use statistical inference and your quantity of interest is reasonably lognormaly distributed, then lognormal calculators such as expostats and IHSTAT can be used.

If your model is normal rather than lognormal, you can either exponentiate your measurements and use expostats or use the webexpo algorithms / prototypes which include the normal model.

Curious to hear colleagues with more experience in vibration/heat.