Advanced Statistical Analysis of
Sampled Data
The Measurement Data Entry Worksheet
contains a variety of options for analyzing data samples to estimate
measurement uncertainty due to repeatability error. These include analyzing measured values, measurement data recorded
in sampled cells, or measurement data recorded as sampled mean values and
standard deviations. Mixed samples, consisting of sample mean values with specified sample sizes
together with estimated population standard deviations and degrees of freedom,
can also be analyzed.
Data Entry Options
Up to 2,000 sampled values can be entered or imported into the Measurement Results
table. The
Measurements Results data entry table changes depending upon which sampling
option has been selected.
The
Sampled Values option is selected when entering individual
repeat measurements. The data can
be entered as measured values or deviations from the nominal or specified
value of the parameter.
The Sampled Cells option is selected when entering a sampled value that
has been observed one or more times. In
this case, the value can be entered as a measured value or a deviation from
nominal.
The corresponding number of times that the value was observed is
entered under the Number Sampled column.
The Sampled Mean Values option is selected when entering mean values
obtained from sets of repeat measurements.
The mean values or deviations from the parameter nominal value are
entered along with the standard deviation and sample size for each set of
repeat measurements.
The
Mixed option is selected when statistics from a prior sample analysis
will be used to characterize the repeatability uncertainty of a measurement
process. The recently collected measured data are entered on the
Measurement Results table and a Mean Value is computed and displayed in the
Computed Statistics portion of the Worksheet. The standard uncertainty
from the prior sample analysis is entered into the Distribution Uncertainty
Estimate field and the associated sample size is entered into the Applicable
Sample Size field. With the
Mixed option, we use these variables to represent the repeatability of the
measurement process and to estimated confidence limits for the computed mean
value with a specified confidence level.
The Data Import
Profile Screen, which is
accessed by selecting the Import Data option on the Tools menu, can be used to import data directly into
the Measurement Results table from the following external application
files: Microsoft Access, dBASE
III or IV, Microsoft Excel 3.0 through 2000, Lotus
1-2-3 WK1 through WK4, HTML and delimited text
files.
Computed Statistics
The
computed statistics also vary slightly depending upon which Sampling Option
has been chosen. The computed statistics for all four sampling options include
the Sample Size, Sample Mean, Mean Deviation, and Standard
Uncertainty. The user can
input a Confidence Level (%) or Coverage Factor to compute the Confidence
Limits. The
display precision (e.g., 1 decimal place, 2 decimal places, ...) for the
computed statistics can also be set by the user.
The Coefficient of Skewness and Coefficient of Kurtosis are also
computed for the Sampled Values and Sampled Cells options.
Skewness is a measure of the asymmetry of the sample distribution.
Kurtosis is a measure of its "peakedness." The
Within Sample Sigma and Between Sample Sigma statistics are also
computed for the Sampled Mean Values option.
The Between Sample Sigma is the
standard deviation of the mean values and the Within Sample Sigma is an
indicator of the variation within samples.
Additional statistical analysis tools include
Normality
testing and Outlier Removal testing, which are accessed via the Tools
menu at the top of the Worksheet.
Normality testing is used to determine if the
sampled data can be assumed to be normally distributed.
Outlier identification is done using a combination of statistical
testing and the application of Chauvenet’s criterion.
The Use Uncertainty in the Mean option box is checked if the
uncertainty is to be reported for the mean value (or overall mean for
the sampled mean values option).
If this option box is left unchecked, then the standard uncertainty
applies to individual measured values The standard uncertainty
for the mean value is computed somewhat differently
depending upon the Sampling Option chosen.
In some cases, only a single measurement or deviation may be
available for entry into the Measurement Results table.
If so, then a standard uncertainty will not be computed.
However, if a standard uncertainty has been obtained elsewhere
that characterizes the repeatability uncertainty in a measurement
process, it would be beneficial to be able to enter it into the Computed
Statistic section of the Worksheet, along with the appropriate sample
size. But, if data are entered into this
section of the Worksheet, UncertaintyAnalyzer responds by deleting the
measurement sample information in the Measurement Results table.
To prevent this from happening, the Override Sample Size and
Uncertainty option box must be checked.
Sample Histogram
A Sample Histogram plot, located in the upper right portion of
the Worksheet, displays the relative frequency of the
sampled data values. The histogram Auto-cell feature automatically
determines the optimal number of histogram cells to display for sampled
data. Alternatively, the user can enter a preferred number of histogram
cells for plotting.
Correction Factors
If a bias (i.e., systematic offset) from the
parameter's nominal value has been input via the Parameter
Bias Uncertainty Worksheet or Parameter
Template, then the user can select whether or not to include the Parameter
Bias Correction factor to the computed mean value.
A fixed offset in the subject (or measuring) parameter value or reading can
also result from environmental factors (e.g., temperature, humidity, air
pressure, ...). The user can
select whether or not to adjust the computed mean value by an Environmental
Correction. The environmental
correction is estimated using the Environmental
Factors Uncertainty Worksheet, which is accessed via the Environmental
Factors button on the Main Screen Error Source List.
|