|
| |
ARM Dose-Response Analysis Report
|
back |
|
|
|
Introduction-Options
|
|
The dose-response analysis calculates an estimate of the chemical dosage that provides the desired response level. ARM includes algorithms to estimate dosage using either probit or logit analysis.
ARM provides three different Dose-Response estimation
*algorithms: Probit - Least Squares, Probit - Maximum Likelihood,
and Logit. The
estimation method, analysis type, and output format can be
customized on the Dose-Response Analysis Report Options Dialog. |
 |
|
top
|
|
Step 1-Treatment Selection
|
| ARM will always prompt for
the treatments to include in the report. Select only the treatments
that you wish to include in the analysis
NOTE: Because
the Dose-Response Report requires extensive user interaction, it
cannot be run as a batch process.
top
|
|
Step 2 -
Data Column Selection
|
|
ARM always prompts for the data columns to include in the
report. |
|
|
|
|
When determining
which data columns to select for the Dose-Response Report keep in
mind the following points:
-
Select
only the data columns that you wish to include in the analysis.
-
Do
not select data columns that are marked as non-analyzable.
If a non-analyzable data column is selected, ARM will not
include that data column in the analysis.
-
Do
not select data columns that contain sample size information.
Select only the data columns that contain response
information. If the
user selects a data column in this step and then selects the
same data column during the sample size verification step, ARM
will issue a warning and the analysis process may need to be
restarted. See Step
5 - Sample Verification for further discussion of sample size
data columns.
If no valid data columns are selected, the analysis process
will abort.
top
|
|
Step
3 - Chemical Selection
|
|
ARM attempts to determine which chemical to include in the
analysis by scanning the chemicals present in the selected
treatments. When ARM is unable to automatically determine which
chemical to analyze, ARM displays the Chemical Selection Dialog.
The user should select the chemical to use in the analysis
from the list of available chemicals.
NOTE: Treatments
that are marked as untreated on the Settings Dialog Treatment Tab
are used as control treatments in the analysis and will not be a
part of this treatment scanning process.
-
If
there is one chemical that is present in each treatment, ARM
automatically selects that chemical.
-
If
there is more than one chemical that is present in each treatment,
ARM must prompt for the chemical to use.
-
If
there is not a chemical that is present in every treatment, ARM must
prompt for the chemical to use.
top
|
|
Step
4 - Rate Verification
|
|
After the chemical to analyze has been determined, ARM allows
you to view and edit the selected rates. In some cases the selected chemical may be present more than
once in a treatment. ARM
defaults to the first occurrence, but the user may select the
desired occurrence from a list of rates and corresponding
application codes on the Rate Verification Dialog.
If the selected chemical was not found in a selected
treatment, ARM fills the rate with "Not Present".
This provides you with a chance to fill in any desired rate.
|
|
|
|
IMPORTANT NOTES: It is the responsibility of the user to insure that the rates
listed use a consistent rate unit.
To assist the user, ARM lists the original rate unit for
reference. At this
point, the user should be sure that the rates are consistent.
Changes made on this screen will not be transferred back to
the treatment editor.
-
All
treatments with a rate of "Not Present" will be
dropped from the analysis.
-
All
treatments with a rate less than zero will be dropped from the
analysis.
-
All
treatments with a rate of zero will be treated as a control.
-
If
there is more than one control treatment in the selected
treatments, corresponding data will be averaged or summed
depending on the data column type (percent will be averaged,
response will be summed. See Step 7 - Data Preparation for
further explanation).
-
At
least 2 valid treatments/rates are required (not counting
controls) for the analysis process.
top |
|
Step
5 - Sample Size
|
|
Sample Size is required information for the dose-response
analysis process. ARM
provides two different ways of specifying sample size.
The Sample Verification Dialog Sample Size option allows the
user to select "Use Fixed Sample Size" if the sample size is
the same for all observations or "Use Sizes in Data Column
#" if sample sizes differ among observations and the sample sizes
to use are entered in a data column.
|
|
|
|
Note: The
sample size data column may not be one of the data columns
selected in step 2. If
the sample size data column specified on the Sample Verification
Dialog has already been selected in Step 2, ARM will display the
following warning:
|
|
|
|
|
In order to proceed, a different sample size data column
number should be specified. If
the specified sample size data column number is correct but has
already been included in the selected data columns, the report
process must be restarted. Be
sure to select only response data columns in Step 2.
ARM must also verify the type of assessment data contained in
the data columns that will be analyzed.
Select "Percent" if assessment data is entered as
percentages. Select
"Response" if assessment data is entered in raw units.
top
|
|
Step
6 - Data Preparation
|
|
A dose estimate will be computed for each valid selected data
column. (If the
analysis is simple, "Simple" is the selected Analysis Type
on the Dose-Response Analysis Report Options Dialog, then only one
data column will be analyzed.)
ARM will determine a sample size, response, and percent kill
for each valid treatment rate in the data column. There are four different scenarios from which ARM must
process these values. Sample
size may either be a fixed number or may be read from a data column. The data columns may either be "percent" or
"response" data columns.
These options are set on the Sample Verification Dialog as
discussed in Step 5. The
example below describes the four different calculation processes
that ARM may use to prepare the analysis data.
|
|
|
|
|
|
After ARM determines a sample size, response, and percent
kill for each treatment rate, ARM validates this information
according to the following checks.
Data that can not be validated is dropped from the analysis.
Percent must be greater and 0 and less than 100.
For the Logit Estimation method and the Probit Least Squares
Estimation method, the ARM performs the Berkson's Adjustment on
percents of 0 and percents greater than or equal to 100.
top
|
|
- Example: Fixed Sample Size with Response Data Column
|
|
|
|
If sample size has been specified as a fixed number and the
selected data columns have been identified as "response"
data columns ARM will calculated sample size, response, and percent
based on the following formulas:
-
Response = The sum of all assessment data for the current
treatment in the current data column.
-
Sample Size = The number of observations summed to get the
Response * Fixed Sample Size
-
Percent = Response / Sample Size
|
|
|
|
Example: Using
Treatment #1, Data Column #1 with a Fixed Sample Size of 50
Response = 40
+ 41 + 42 + 43 = 166
Sample Size = 50 * 4 observations = 200
Percent = 166/200 = 83%
top
|
|
- Example: Fixed Sample Size with Percent Data Column
|
|
|
|
If sample size has been specified as a fixed number and the
selected data columns have been identified as "percent"
data columns ARM will calculated sample size, response, and percent
based on the following formulas:
-
Percent = The
average of all assessment data for the current treatment in the
current data column.
-
Sample Size = The number of observations averaged to get the
Percent * Fixed Sample Size
-
Response =
Sample Size * Percent
|
|
|
|
Example: Using Treatment #1, Data Column #2 with a Fixed
Sample Size of 50
Percent = (80 + 79 + 74 + 70) / 4 = 76%
Sample Size = 50 * 4 observations = 200
Response = 200 * 76% = 152
top
|
|
- Example: Sample Size Data Column with Response Data Column
|
|
|
|
If sample size has been specified as a data column and the
selected data columns have been identified as "response"
data columns ARM will calculated sample size, response, and percent
based on the following formulas:
-
Response = The sum of all assessment data for the current
treatment in the current data column.
-
Sample Size = The sum of all assessment data for the current
treatment in the sample size column.
-
Percent = Response / Sample Size
|
|
|
|
Example: Using
Treatment #1, Data Column #1 with a Sample Size of Data Column #3
Response = 40 +
41 + 42 + 43 = 166
Sample Size = 45 + 45 + 50 + 48 = 188
Percent = 166/188 = 88%
top
|
|
- Example: Sample Size Data Column with Percent Data Column
|
|
|
|
If sample size has been specified as a data column and the
selected data columns have been identified as "percent"
data columns ARM will calculated sample size, response, and percent
based on the following formulas:
-
Percent = The
average of all assessment data for the current treatment in the
current data column.
-
Sample Size = The sum of all assessment data for the current
treatment in the sample size column.
-
Response =
Sample Size * Percent
|
|
|
|
Example: Using Treatment #1, Data Column #2 with a
Sample Size of Data Column #3
Percent = (80 + 79 + 74 + 70) / 4 = 76%
Sample Size = 45 + 45 + 50 + 48 = 188
Response = 188 * 76% = 143
top
|
|
Step
7 - Time Series Interval Verification
|
|
|
|
If "Time Series" is the selected Analysis Type on
the Dose-Response Analysis Report Options Dialog, ARM must determine
an interval for each valid data column.
ARM first looks at the Trt-Eval Interval Field
("TEI"). If
there is an entry, ARM will use that for the default interval.
If there is not a Trt-Eval Interval, ARM will next look to the
Rating Date Field ("EED").
If there is a rating date entered, ARM will use that date in
reference to the first application date found in the site description
("ADx").
ARM will convert the intervals to the lowest common
denominator.
top
|
|
Step
8 - Analysis Graph
|
|
ARM allows the user to view and customize the graph prior to
printing. There are two
types of graphs that can be displayed, simple and time series, depending
on the selected Analysis Type on the Dose-Response Analysis Report
Options Dialog .
A Simple Analysis is performed on one data column.
The Simple graph plots the original Percent vs. Rate data used to
calculate the dose estimate. ARM marks and labels the calculated dose estimate with a
hollow point and the estimate abbreviation and percentage entered on the
Dose-Response Analysis Report Options Dialog.
ARM draws a probit or logit line based on various LCp points.
The graph includes confidence level bars if "Confidence
Levels" is selected on the Dose-Response Analysis Report Options
dialog.
Note: For
a probit estimate method, the probit line drawn corresponds directly to
the slope and intercept equations that may be printed on the report.
However, when using the logit estimate method, the logit line
drawn is not based on the slope and intercept equations on the report.
A Time Series Analysis is performed on multiple data column.
ARM calculates a simple dose estimate for each selected data
column. ARM determines a
time interval for each data column according to the algorithm discussed
in Step 7 - Time Series Interval Verification.
The Time Series graph then plots each dose estimate/time interval
point and connects these points with a spline curve.
top |
|
Acknowledgements
and References
|
|
*Algorithms were provided courtesy of Dr. J. J.
Hubert, University of Guelph.
Other References:
Hubert, J. J. BIOASSAY. Dubuque: Kendall/Hunt
Publishing Company, 1992.
top
|
|
