Project Name Industrial Hygiene Monitoring Introduction To Bayesian
Project Name Industrial Hygiene Monitoring Introduction To Bayesian Statistics Dave Pearson, Djon Gentry, Tim Allers 1
Project Name INTRODUCTION TO BAYESIAN STATISTICS 2
Project Name Introduction to Bayesian Statistics What is Bayesian Statistics? – An approach to statistics in which estimates are based on a synthesis of a prior distribution and current sample data. – Belief that something will happen is based on past observation. 3
Project Name Introduction to Bayesian Statistics What is Bayesian Statistics? – Everyday Example – When you cross a road, you may need to do a full "analysis" of the data concerning oncoming traffic (How fast? How big is the gap? Does the driver look like a maniac? Have they seen me? ) but one thing you don't consider: once the car has gone past and I step out on the road; what is the chance that it will back-up and run me over – In essence we've performed a Bayesian calculation based on a lifetime of crossing the road. We have a strong belief, based on experience, that the car will not back up. 4
Project Name Introduction to Bayesian Statistics Why use Bayesian Statistics? – Quantifiable validation of prior and future PPE & engineering control decisions – To predict future exposures and establish PPE and engineering control requirements – Typical employee works 220 days/year. 4 samples represents roughly 2% of their annual exposure profile – Statistically valid decision making with as few as 3 samples – New EHS Standard requires that a statistical tool be used – Provides a consistent process for IH decision making – Risk Management Decision making based on defensible data 5
Project Name From: Paul Hewett, Ph. D. , CIH & John Mulhaussen, Ph. D. , CIH 6
Project Name From: Paul Hewett, Ph. D. , CIH & John Mulhaussen, Ph. D. , CIH 7
Project Name From: Paul Hewett, Ph. D. , CIH & John Mulhaussen, Ph. D. , CIH 8
Project Name From: Paul Hewett, Ph. D. , CIH & John Mulhaussen, Ph. D. , CIH 9
Project Name From: Paul Hewett, Ph. D. , CIH & John Mulhaussen, Ph. D. , CIH 10
Project Name From: Paul Hewett, Ph. D. , CIH & John Mulhaussen, Ph. D. , CIH 11
Project Name Bayesian Decision Analysis Professional Judgment * 2 Trivial Highly-Controlled Well-Controlled Poorly-Controlled * Qualitative Exposure Assessment Input 12
Project Name Opening & Running the IHData. Analyst Software 13
Project Name Starting the IHData. Analyst Program On your desktop click the IHData. Analyst icon. Either this Or This 14
Project Name The Program will Launch 15
Project Name Input Facility Information Here The facility information should reflect the area where the Industrial Hygiene monitoring data was obtained. 16
Project Name Input the Substance Information Here. This section would include the chemical name or if the monitoring was for noise, the word “noise” Next, click the drop-down box and select the unit of measure. 17
Project Name Input IH Data Here Input the monitoring data for the Substance monitored in the “Conc” column. Place a “Y” in the “LOD” column if the data is a less than (<) value. 18
Crunching the Numbers Click the “Calculate All” button on the toolbar. Project Name 19
Project Name IH Data Input & Decision Making 20
Project Name There’ s a lot of Stats in there! Help me sift through it. XO. 95 = 95% of your data should fall below this point. 95%UCL = 95% confident that any new data will fall below this number based on existing data. BDA Charts – Indication of how well controlled your process is. PPE Charts – J&J IH has agreed to use “ 75% or >” for deciding if PPE is warranted and the appropriate APF. 21
Project Name Example 1 Ethinyl Estradiol All samples less than half the OEL 22
Project Name Example 1 All 5 samples < 0. 005 XO. 95 23 95% UCL
Project Name Example 1 24
Project Name < 75% What Is Your Decision? 25
Project Name Removed “<“ Decision Probability Increased 26
Project Name DON’T FORGET TO CALCULATE ALL!!! Added one more data point. Now > 75%. Feeling more comfortable with Making decision for no respirators 27
Project Name By doubling the original data Resulted in 28
Project Name 29 Doubled the data. Greater comfort in PPE decision Now 84. 5%
Project Name Report Format What does this program provide as far as from a written document? 30
Project Name Report - Example 1 31
Project Name Report - Example 1 32
Project Name Report – Example 1 33
Project Name Report – Example 1 34
Project Name Example 2 Cobalt dust/fume Sample results at or less than OEL Look at your data. What is it trying to tell you? Knowing when to take additional samples and knowing when to stop sampling and implement controls. 35
Project Name 8 data points ranging From 1/10 to 1 X OEL 36
Project Name BDA Charts = 4 Poorly Controlled 37
Project Name PPE Charts = 10 X 38
Project Name Decision Making BDA Charts indicate Poorly Controlled PPE Charts indicate the need for 10 X APF What are your next steps? Make sure employees performing this work are included in the RPP, Med Surveillance, and provide a minimum of 10 x protection respirators. Do you think that obtaining additional samples will help at this point? • Probably Not at this time Enhance engineering controls, then re-sample • After engineering controls have been implemented, new data collected should be separated from the previous data collected during Poor Controls 39
Project Name Example 3 8 Hr TWA vs. Task Duration You obtain samples during a specific task The task lasts approximately 30 minutes for the day Confirmed no other exposure to this or similar compounds the remainder of the day Compound being measured does not have acute hazards associated with it and there is no assigned STEL. What decisions are you making based on the sampling results? 40
Project Name Example 3 – 8 Hr vs. Task All results are from sampling 30 minute task Results are 1 to 3 X OEL (No STEL) 41
Project Name Example 3 – 8 Hr vs. Task PPE Charts indicate the need For employing 10 X APF 42
Project Name Example 3 – 8 Hr vs. Task However, if one were to convert results from lab into 8 HR TWA, assigning ZERO remainder of the day 43
Project Name Example 3 – 8 Hr vs. Task The PPE Chart now indicates No Respiratory Protection needed 44
Project Name What Decisions Are You Making? Are you “In Compliance”? Is compliance the only thing we should be looking at here? What are you doing in the absence of an STEL? Are you confident that the compound has no acute effects? Does your compound have an R-Sen D-Sen notation on the MSDS? 45
Project Name Can We Make A Decision With Just 3 Samples? 46
Project Name Can We Make A Decision With Just 3 Samples? 44% Confident with 10 X 85% Confident with 25 X 97% Confident with 50 X 47
Project Name QUESTIONS? 48
- Slides: 48