Reports
Soil data may be viewed in three different ways – Summary of All Labs, Single Lab View, and a list of trials whose result exceeds three standard deviations. Let’s refer to this last view as the Error View, for convenience. The Error view will take on two forms.
1) A list of all “problem” results. This view is displayed when a Summary of All Labs is being viewed.
2) A list of “problem” results for a single lab, displayed when data of only one lab is being viewed.
Controls
Click the Reports link and then log in.
Logging into the Reports application will present you with five additional controls. These controls help you select the report you want to view.
1) Year Combobox – click on the small down arrow to change the year.
2) Month Combobox – click on the small down arrow to change the month.
3) Report Type Combobox – click on the small down arrow to change the report type from Summary to Division 1, 2 and so on.
4) Method – A, or C. (May be removed in future versions).
5) Build Report Button – click this button to process the report.
LOGOFF = this is a new link for this version of STDC. Click this button to change the current user. The Login form will be displayed when you do this.
The report "controls" tell STDC which data to process and then display. As you can see, reports are filtered first by year, followed by then month, and by type.
There are two report types – 1) Summary, and 2) Lab. These reports are described in their own sections below.
Here’s what the form will look like if you click the Build Report button. This may take several seconds when Summary has been selected. So, please be patient.
The first thing to notice is that the title changes to tell you what you are looking at. In this case we have created a Summary Report for January 2003. Summary Reports include trial data from all labs for that year and month.
Summary Reports
In the top view is a tabular presentation of all tests run on the samples. It summarizes the results for each test by calculating the mean (or average), the standard deviation (SD), and the coefficient of variation (CV%). Coefficient of variation is a measure of the distribution of results about the mean. A small CV implies that most of the results lies clusterred close to the mean, while larger values of CV imply data that is more spread out.
To help with the analysis the bottom portion of the report presents a listing of trials whose tests results exceed 3 standard deviations. This error report displays the information you need to find and analyze unusual trial results. The error report lists the lab, test, and trial in question followed by the result recorded, the difference from the mean, and the standard deviation for all 30 trials.
Lab Reports
Changing the Report Type Combo from Summary to one of the labs, in this example Division 6. We get a different kind of report. This new report is simply a listing of the trial results for that lab on samples for a year and month. In this example Division 6 – 2003 – January. In other words, this report shows the trial results for samples from Division 1 in 2003.
The error report for the Lab View indicates trials with unusual results (i.e. results exceeding 3 SD). By looking at trial results in the Lab View we see that trials 1 and 3 were much closer to the mean and therefore we could conclude that this result is in error. The latest version of STDC also includes lab and test ratings for each trial.
Calculation Details
Mean is a calculated by adding all trial results and dividing by the number of trials. In our case, each lab conducts 3 trials for each test. That means the total number of trials will be 30 for any given soil sample and any given test. STDC does not take this for granted. Instead it counts the number of trials it finds and used this count for its divisor.
As a check we will manually list and calculate the mean for 2003 – January – Passing 2.36 mm Seive.
X[k]
89.2
87.0
84.0
72.1
73.5
84.1
89.5
85.4
85.4
79.0
80.0
79.0
73.9
73.9
73.9
81.0
84.0
84.0
77.7
58.7
73.8
70.0
70.0
70.0
74.3
74.3
74.3
64.0
63.0
65.0
Total = 2292
Mean = 2292 / 30 = 76.4
Standard Deviation – for a set of results we can expect the dispersion about the mean (the difference from the average) to be predictable such that values with a dispersion of more than 3 standard deviations are considered unusual and worth further investigation. Standard deviation can be calculated with the following formula.
SD = Square Root of ( Sum of Differences Squared / The number of trials – 1)
In computer ease we use this:
Xbar = mean
For k = 0 to k < TotalTrials
SSQ += (X[k] - Xbar) * (X[k] - Xbar)
Next
SD = sqrt(SSQ/(TotalTrials – 1))
SD = 8.11
CV% = (SD/Xbar)*100
CV% = (8.11/76.4)*100 = 10.61%
For reference: 2*SD = 16.22, 3*SD = 24.33
Difference used in the Error listing is abs(Xbar – X). So, for our 30 trials we have,
Abs (Xbar - X[k])
12.8
10.6
7.6
4.4
2.9
7.7
13.1
9
9
2.6
3.6
2.6
2.5
2.5
2.5
4.6
7.6
7.6
1.3
17.7
2.6
6.4
6.4
6.4
2.1
2.1
2.1
12.4
11.4
11.4
All of these differences fall within 3*SD therefore no errors will be reported. If one of these results had been larger than 3*SD then an error would have been reported.
References
1) Web reference for mean, standard deviation and coefficient of variation can be found at this site:http://www.westgard.com/lesson34.htm
2) Any college statistics text.
