Demonstration of Monte Carlo Software

1
Demonstration of Monte Carlo Software
2
Note to the user

• This presentation should be viewed as a slide
  show.
• Whenever the user sees the symbol , they
  should click on the left button of the mouse.
  This will bring up arrows, screen shots and
  guidelines to help the user.

3
Introduction

• This demonstration
• shows the main features of the software.
• demonstrates how the Monte Carlo software has
  been
• specially designed to be used by food chemical
  exposure
• assessors

4
Logging in

• First you will need to apply for an account. To
  do this click the Register button.
• Once you have been sent an email with a username
  and password you can log in.

5
Planning

• Users can use expressions (distributions) to
  describe their input data
• If you are not quite sure how to describe your
  input data using a distributuion you can look at
  some of the options available in the List of
  Expressions.

6
Testing expressions

• Looking at the lognormal, you might decide this
  is the expression you want.
• From here within the Expression Tester you can
  input this expression.
• If you are interested in body weight, and your
  data has a mean of 70kg and a standard deviation
  of 10, you type this in as shown in the screen
  shot and click submit.

7
Results of expression testing

• Clicking submit will give a results page as
  follows
• To view the histogram for this distribution
  simply click on the Histogram link.
• A histogram such as this will appear.

8
Creating tables and uploading data

• Creating tables and uploading data is organised
  through the database manager.
• The type of databases
• you might need to create and upload include
  databases on food consumption, recipes, chemical
  concentrations etc.
• To create a new table click here

9
Example table

• Creating a new table will show you the following
  screen.
• All table types included in the sofware are
  located in the drop down menu here.
• In this example a food diary table is created, so
  diary should be selected from the drop down
  menu and a name and description filled in.
• Then click Go

10
Requirements

• You will need to specify the fields for the table
  and the type of data you will be entering. For
  example in the diary table there is subject code,
  survey day, time, meal number, food code, brand
  code, amount consumed and weekday
• The amount field has the field type set as an
  expression. This means that you can enter
  data either as a fixed value or as a
  distribution/expression

11
Customisation

• Example you want to look at intakes amoungst
  males and females separately. For this you may
  want to add a variable, gender to you table
• You would use an integer because males and
  females are usually coded as a 1or 2
• You could call the variable Gender and say it
  refers to gender of subject
• Then click go to add this field
• Click install when you are happy you have all
  the fields you need. This creates the table
  structure to which you are going to upload data.

12
Uploading databases

• A nice feature of the software is that the data
  which you upload can come from lots of different
  sources (e.g. excel spreadsheet, oracle database,
  text file etc.) and it can come in any language
  once the fields are in the order that has been
  specified in the table.
• Here, if you have a text file which you are going
  to upload you click the upload link beside the
  relevant table.
• You will need to click Browse to locate the
  file on your computer. Then click Upload File
  once you have found the file to be uploaded.

13
phpMyAdmin

• One way of viewing your uploaded data is by using
  phpMyAdmin. (a web
  interface for a database management structure)
• Clicking here will bring up this screen.
• The user should then choose the datbase they wish
  to view (e.g. demo).
• Choosing a database like this will bring up a
  list of tables in this database. The user
  can then Browse each individual table.

14
Data and databases

• You can then proceed to install and upload other
  tables relevant for your exposure assessment
  (e.g. chemical concentration, subjects etc.)
• One important thing to mention is that the
  software has been designed to include all the
  types of databases that you would ideally have in
  exposure assessment for food chemicals.
• In the real world, you may not have all of these.
  For example you may not have a database of
  processing factors or recipes.
• This does not mean you cannot use the software.
  You still have options to enter conservative or
  default values for variables or distributions.
  And for certain non-critical fields in the
  tables, the fields can be just left blank.

15
Food groups

• If you are doing an exposure assessment, you have
  to match food consumption data with chemical
  concentration data. this example shows you how to
  make a food group.
• Here, for example, you can make a food group for
  margarine. You fill in Margarine for the food
  group name and select the table you wish to use
  from the two drop down menus.
  Then click Go.

16
Populating food groups

• If you know the exact food codes you can type
  them in directly.
• Or you can search a foods database for the text
  string Margarine (retrieves food codes with
  margarine in the name).
• Or you can carry out an advanced search (to
  include all foods which contain margarine as part
  of a composite dish).
• Note in order to perform these searches a foods
  database and a recipe database will need to be
  installed an uploaded through the database
  manager.

17
Submitting jobs

• When you have created and uploaded all the tables
  relevant for you exposure assessment, you can
  click on Submit Jobs
• And from here you can click the link to set up
  and run a new simulation.
• This brings up a screen with the three types of
  job available to the user. For this example, a
  basic chemical simulation is chosen.

18
Simulation tables

• You pick the tables containing information for
  -subjects
  -food consumption
  -chemical concentration.
• If you like, you can select subsets of data from
  within the tables. E.g you may only want to look
  at subjects with a body weight gt 55kg.

19
Simulation options

• Now you are going to set your simulation options.
• If you have small sets of data or if you want to
  model food consumption data parametrically (e.g.
  using a lognormal distribution), then you need to
  use the subject replication option.

20
More simulation options

• The next option is bootstrapping. You can
  bootstrap on subjects, on days or on eating
  events. You need to specify how many bootstrap
  sets you want.
• You can choose a number of iterations for your
  simulation.
• You can choose the percentiles at which you want
  to view exposures.
• Finally when everything is ready, click Submit.

21
Computation

• After clicking Submit, the job enters a queue
  on the Monte Carlo software and results are
  calculated using high performance computing.
• You can view the status of the job at any stage
  and see what stage it is at in the simulation
  process.

22
Tracking

• You will notice that the job was given an ID
  number so if at any stage you want to see the
  simulation setup you can go to the results
  section.
• From there you can click on Job Summary for
  that ID number.
• This will bring up the following summary page.

23
Viewing Results of a Simulation

• To view results click on View Results and
  then the Results link of the job you wish to
  view
• A summary page similar to this appears.
• To view the simulation results over the
  reference set (across all iteration) click on
  this link.

24
Visualising results

• On this page is a table for producing a
  customised graph.
• In this case the 95th percentile for the total
  population (all_p95) is chosen along with the
  chemical intakes per kg body weight per day (cba)
  and a regular histogram. If you want to look at
  the probability of exceeding the ADI (in this
  example 0.005) you can type in this value.
• The resulting histogram will then be produced and
  will look like this

25
Boxplots

• You can also look at the results in the form of a
  boxplot.
• In this case the 99th percentile for consumers
  only (cons_p99), and for the average daily
  chemical intake per kg body weight (cba) is
  chosen.
• The resultsing boxplot looks like this.

26
Downloading data

• You can also download the results data in a
  number of forms. You can download results as a
  html file and view them on screen or you can
  download them as txt or dif files which can
  be opened in other programmes such as Excel.
• An example of results in HTML format is shown here

27
Viewing specific data

• If you have a particularly high chemical/food
  intake, you may want to identify which subjects
  these hight intakes relate to.
• By downloading and sorting the data table you
  can see which iteration contained the high value.
  You can then ask to view the raw data for this
  iteration.
• Looking at this raw data you can see the pattern
  of food consumption and chemical concentration
  for every subject which led to each chemical
  intake within this iteration.

28
Example

• For example, for subject number 1 his/her intake
  of the chemical from food code 310 came from
  consuming 46g of the food which had the chemical
  present at 1.29mg/kg and he/she had a body weight
  of 79.6kg. This give an intake of 0.0007.
• You can store and view this very detailed
  information for your exposure assessments so any
  unexpected values can be checked.

29
Parametric distributions

• As mentioned earlier you can enter food
  consumption and chemical concentration values
  directly as a parametric distribution, in
  addition to raw data.
• Looking at the food consumption table,
  Diary_annatto_parametric, which was uploaded in
  database manager, you see that intakes are
  expressed as the product of a Bernoulli
  distibution (probability of being a consumer or
  not) and a lognormal distribution (describes
  level of intake).

30
Example

• It is possible to look at food group 387,
  which is spreadable fats, in the expression
  tester
• This distrbution gives the following histogram.
  You can see there is a peak of zero intakes for
  non-consumers (about 9).
• The rest of the histogram represents the intakes
  among consumers with a mean of 18.8 and a
  standard deviation of 18.0

31
Food intake simulation

• You can now use the Diary_annatto_parametric
  table in a food intake simulation.
• In the job set up tables are selected for
  subjects and food consumption. When you get
  to simulation options, because the food
  consumption is a parametric distribution, you
  need to replicate the subjects. Any number of
  iterations can be set.
• As usual, it is possible to view the results of
  this job using either boxplots or histograms.

32
Comparing jobs

• It is also possible to compare the results of up
  to ten simulations.
• For example, one job could have food intake
  entered as raw data (job 1842) and another with
  food intake entered as a lognormal distribution
  (job 1843).
• These ID numbers are entered in the text boxes on
  the main results page and Go is pressed.

33
Results of comparing jobs

• The resulting page will give a form like this
  where you can choose the type of graph you would
  like to view.
• Here the distribution of 95th percentiles for the
  total population (all_p95), for average daily
  chemical intakes (caa), and the reference set
  (i.e. across iterations) of each job is chosen.
• A graph of the two boxplots will be produced.

34
Extra simulation options

• If full chemical simulation options is chosen
  instead of basic chemical simulation, a number
  of extra options are displayed for the job set
  up.
• Example, when you are selecting the pesticide
  concentration data to use in your model, you may
  have lab data for some pesticide crop
  combinations but not for others. Where it is
  missing you might want to use the maximum residue
  level (MRL). This option allows you to specifiy
  if you want to use lab data only, MRLs only or to
  only use MRLs for food groups that there is no
  laboratory data for.

35
Limit of reporting (LOR) model

• One major consideration for pesticides is
  deciding how to treat pesticide residues that are
  below the limit of reporting.
• Analytical methods may not be sensitive enough to
  detect a residue even if it is present.
  Therefore, a non-detect or value below the LOR ,
  does not mean no residue is present. You might
  want to consider a number of scenarios
• 1) You might want to treat values below the LOR
  as true zeros (e.g. where agricultural staistics
  indicate that it was very unlikely that the
  pesticide was used to treat crops).
• 2) You may want to assign values of half the LOR
  to any values below the LOR.
• 3) You may want to assign values of the full
  LOR to any values below the LOR.

36
Using the LOR model

• The software allows you to choose any of these
  three options.
• You can run a model three times, using each
  option, to see does it make a difference to the
  exposure levels.
• The difference that it makes will probably be
  related to the number of samples below the LOR
  and the magnitude of the LOR to the positive
  residue levels.

37
Unit-to-unit variability

• Another issue when modelling acute pesticide
  exposure is consideration of unit-to-unit
  variability in composite samples.
• For example, you do not sample a single apple,
  you sample a box of apples and the effect that
  one bad apple will have will depend on the size
  of the box.
• The partners in the Monte Carlo project developed
  a number of methods, taking into account factors
  such as shared history and number of units in the
  composite sample to take account of this
  unit-to-unit variability (for more detail please
  refer to Users Manual).

38
Further Reading

• This presentation is only an introduction to the
  software to allow users see some of its
  capabilites and main features. The project
  website is http//www.tchpc.tcd.ie/montecarlo.
• Here you can find
• 1) more information on good modelling practice
  for food chemical exposure analysis. Please refer
  to the paper entitled Practical Guidelines.
• 2) more information on the software itself and
  all its features from a users perspective.
  Please see the Users Manual.