Sie sind auf Seite 1von 11

Chris Harding

Statistical Analysis: Representation Success at Board of Veterans Appeal 1 of 11




Date: 29 August 2013
Statistical Analysis: Representation Success at Board of Veterans Appeal

US Veteran
United States of America
CC: US House Committee on Veterans Affairs; US Senate Committee on Veterans Affairs
To Whom It May Concern:
Purpose
First, the veteran, one who is filing a claim, must be active in the process in my opinion and part of that activity is reading
and becoming informed.
Next, I believe the following statistical analysis will help veterans choose a Veterans Service Organization[1] that maximizes
their success with the claim process. Although I cannot guarantee success, I have used the statistical methods[2-6] of
average, standard deviation, range, maximum, minimum, normal quantile plots and other simple graphical procedures,
Analysis of Variance (ANOVA), and Duncan multiple-range test to increase the likelihood of success by ordering Veterans
Service Organizations according to success, percent allowed when presenting before Board of Veterans Appeals[7]. After
reading this paper, I believe you will, at a minimum, agree that one should not represent themselves before the Board of
Veterans Appeals. Also, the data was extracted from Board of Veterans Appeals Annual Reports[7]. As such, the data does
not, in any way, suggest that a certain Veterans Service Organization will be better than another when dealing with Veterans
Benefits Administration. Rather, the data provides a statistical ordering of Veterans Service Organization success before the
Board of Veterans Appeals. In the future, I hope Veterans Service Organizations will share their Veterans Benefits
Administration successes with me so that I can do a similar statistical analysis.
My VSO Choice
When I had filed a disability claim with the Veterans Affairs, I wanted to maximize my success. Although I had strong
medical evidence, I knew that mistakes could occur. As such, I expected to face an appeal process that might end in the
courts. I have always heard that a fool represents himself or herself in a court of law. Therefore, I believed my success was
dependent upon a good Veterans Service Officer (VSO)What Veterans Service Organization though?
I tried to find Veterans Service Organization comparisons with no success, but I found many social network threads that said
the Veteran Service Organizations were basically the same. Still, this did not make any sense to me because humans run the
organizations and we humans always fall into statistical populations. As such, I suspected the Veterans Service Organizations
fell within populations regarding their success with veteran claims processes. I could not find any comparisons though. As
such, I eventually went with the well-recognized Disabled American Veteran (DAV) organization. Still, I was worried
because my choice was blind.

Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 2 of 11
Board of Veterans Appeals Annual Report to Congress
As a person with two Science, Technology, Engineering, and Mathematics (STEM) degrees: chemical engineering and
biological sciences I have been trained to research most important tasks. Like other veterans, I tried to peak within the
opaque process of the US Veteran Benefit Administration. Still, I found no data regarding VSO success with Veterans
Benefit Administration (VBA) claims. What percent of filed VBA claims were successful and what percent was denied when
considering VSOs? Personally, I still have not found VBA/VSO data. Since I accepted that I might have to appeal, I
eventually discovered the Board of Veterans Appeals (BVA) and the BVA annual reports to Congress. In those reports, I
found raw data, data that has not been manipulated by statistics, that covered the most common Veterans Service
Organizations from 1992 to 2012. To be specific, the table is called BVA Dispositions by Representation FY 2012 in the
2012 Annual report[7]. The BVA reports cases Allowed; Remanded; Denied; Other; and Total.
Raw Data
I decided to create a Microsoft Excel workbook with all available raw data from 1991-2012 representation tables[9]. While
collecting that data, I discovered that the year of 1991 was missing a table, so I collected the year range from 1992-2012.
Some Useful and Informative Graphs
First, I wanted to graph the data and see what I was up against. To do this, I had to rearrange the collected data[9]. Also, I
calculated the average, standard deviation, range, max, and min. After graphing all Veterans Service Organizations (VSO)
representations of percent allowed from 1992-2012, I graphed the means of percent allowed. Interestingly, I found that
most of the represented Veteran Service Organizations (VSO) had a trend of increased success, percent allowed, since
1992. I also noticed that the rate of increase from 1991 to 2012 was variable between representations. One major exception
was Vietnam Veterans of America (VVA), which consistently fell between 19.16% allowed and 43.32% allowed with an
average of 27.33% allowed. In the opposite direction, a representation of None had a slight positive rate from 1992 to
2012, with values between the range of 11.89% allowed to 23.30% allowed and an average of 16.90% allowed. As a side note,
I often I interchange mean with average. Mean and average are the same mathematically.
Now, I provide two graphs and the knowledge that statistical analysis indicated that the averages, second graph, are, on
average, statistically differentANOVA. I describe the ANOVA calculation later. Also, I will eventually order the averages
according to Duncan multiple-range test[2;6]. The averages seen in the second graph are not ordered or grouped according
to Duncan multiple-range test, which will come later. Also, the first graph is all the raw data from 1992 to 2012.
Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 3 of 11





As mentioned previously, ANOVA[2;3;5], which is a statistical tool to compare 2 or more means, indicated that there is a
statistical difference in the percent allowed averages. Also, the reader can see that Vietnam Veterans of America has the
greatest percent allowed average while None has the lowest percent allowed average. Later, I determine which means can
be separated from each other according to Duncan multiple-range test[2;6]. Remember, a mean is mathematically
identical to the average.
What I do not do with the following statistical analysis is determine why one average differs from another. As an example,
a change in the law might have affected the averages.
Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 4 of 11
Analysis of Variance, ANOVA, Results
I am not a statistician, but I am a chemical engineer who took engineering statistics in college, and I used statistics as a
chemical engineer in the pharmaceutical industry to optimize my methylation process for biaxin antibiotic. Still, I am not a
statistician. As such, I will not attempt to reproduce a textbook explanation of ANOVA[2-6].
Before I calculated the ANOVA table, I used Normal Quantile Plots to check if the raw data came from a normal
population[9]. All data, except the Jewish War Veterans, proved to satisfy the normality test. I also compared the standard
deviations of the data[9]. As for standard deviations, the average was 4.93 and the range was 4.71, which is quite variable in
my opinion, but my statistics textbook stated that the following ANOVA method is robust and relatively insensitive to cases
where data does not meet the normality and equal variance criteria[2]. As previously mentioned, all the analyzed data met
the criteria or normality[9]. Jewish War Veterans and American Red Cross were not included in this ANOVA calculation
because their data did not span the 1992-2012 range. Although I could have used another ANOVA method that allows
variable sample size, I decided to exclude the two from my current analysis.
What I will do is provide a very brief explanation of ANOVA, provide an ANOVA table, and provide a comparison to the

F
0.05
distribution table[2;8]. Later, I group the averages of percent (%) allowed according to Duncan multiple-range test.

Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 5 of 11
In a very brief nutshell, the ANOVA analysis calculates an F value. If the calculated F value is greater than

F
0.05
in a
provided statistical table[8], the null hypothesis that all the means are equal can be rejected, and there is a 5% chance of
being wrong. Consider the following table as an example of reported values that we want to calculate the F value. The first
table is not an ANOVA table:

1 (Year) 2 j n Means
Sample 1
(VSO 1)

y
11
(percent
allowed)

y
12


y
1 j


y
1n


y
1

Sample 2

y
21


y
22


y
2 j


y
2n


y
2

Sample i

y
i1


y
i2


y
ij


y
in


y
i

Sample k

y
k1


y
k2


y
kj


y
kn


y
k

Now, I provide the equations used to calculate the F value of the above table[2].

F =
o
B
2
o
W
2
;

o
B
2
= between sample variance;

o
W
2
= within-sample variance
Can be written as:

F =
SS(Tr) /(k 1)
SSE/(k(n 1))
;

SS(Tr) =between-samples sum of squares;

SSE =error sum of squares

SS(Tr) =
T
i
2
i=1
k

n
C;

SSE =SSTSS(Tr);

SST = y
ij
2
j =1
n

i=1
k

C;

C =
T
2
kn

In this paper, k represents the number of Veterans Service Organizations (VSO), and n represents the number of years.

T
i
is the total of the n observations in the ith sample[2]. T is the sum of all

T
i
. The value

y
ij
2
represents the square root of
every percent allowed numerical value, and

represents a summation of all values.



y
i1
+ y
i2
+ y
ij
+ y
in
= T
i
y
k1
+ y
k2
+ y
kj
+ y
kn
= T
k
T
i
+T
k
= T



Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 6 of 11
Once a Microsoft Excel spreadsheet is created, all the above equations can be used to calculate the needed F value[9].
Afterwards, an ANOVA table can be created[2]:

Source of Variation Degrees of Freedom Sum of Squares Mean Square F
Treatments k-1 SS(Tr) MS(Tr) = SS(Tr)/(k-1)

MS(Tr)
MSE

Error k(n-1) SSE MSE = SSE/k(n-1)
Total nk-1 SST
Now, I will provide the ANOVA table I obtained after using Microsoft Excel to organize data and calculate the F value for
the Board of Veterans Appeal percent allowed.

Source of Variation Degrees of Freedom Sum of Squares Mean Square F
Treatments 11 1559.49 141.77 5.44
Error 240 6252.12 26.05
Total 251 7811.61

From the

F
0.05
table[2;8], I get four values. Because of the large degrees of freedom for error, I cannot interpolate between
the last numerical value and the infinity symbol. As such, I report all four values. As you will see, the above-calculated F
value is much larger than the four table values. If I would have interpolated the statistical table, the interpolated

F
0.05
result
would have fallen within the four table values that I provide.

F
0.05
Table
Denominator
Numerator
10 12
120 1.91 1.83
Infinity 1.83 1.75
As the reader can see, the calculated F value of 5.44 is greater than all the above

F
0.05
values: 1.91; 1.83; 1.83; 1.75. Because
the calculated F value is larger than

F
0.05
, it can be assumed that there is statistical significance that the means are different
and not the same. In other words, the null hypothesis of equal means can be rejected.
Now that it is known that the means are statistically different, the Duncan multiple-range test will be used to determine which
means can be separated from each other and grouped statistically. Remember, the words mean and average are identical
mathematically.
Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 7 of 11

Duncan Multiple-Range Test
When performing the Duncan multiple-range test, ranges of means are tabulated. Ranges are calculated from the population
means, Average of % Allowed, and then compared to a test value labeled

R
p
. If the calculated range value exceeds the

R
p
value, the differences in that set are significant and will not be grouped. In this case, we calculate the

R
p
for an

o = 0.05,
which means that there is a probability of 5% to commit a type I error. A type I error is when a rejection occurs even though
the value is true. In this case, there is a 5% error probability that the groupings are wrong.
Raw Data
p r
p
p R
p
Number Representation
Average of %
Allowed
2 2.77

2 3.09 12 None 16.90
3 2.92 3 3.25 11 Agent 20.39
4 3.02 4 3.36 10 State Service Organizations 20.65
5 3.09 5 3.44 9 Veterans of Foreign War 20.88
6 3.15 6 3.51 8 American Legion 21.46
7 3.19 7 3.55 7 Attorney 21.83
8 3.23 8 3.60 6 Other 22.94
9 3.27 9 3.64 5 Disabled American Veterans 23.08
10 3.29 10 3.66 4 AMVETS 23.49
3
Military Order of the Purple
Heart 23.64
2 Paralyzed Veterans of America 24.82
1 Vietnam Veterans of America 27.33

The p is a numerical number from 2 to 10 in a statistical table[2]. The p represents the size of the group that is used to
calculate the range. If I chose the above Representation as an example, a group of p = 3 would be calculating the range:
(Vietnam Veterans of America Military Order of Purple Heart) = 3.69. As can be seen in the next table, the latter range is
statistically significant.
Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 8 of 11
As an example and an actual grouping:

Group 1 Averages: Highest Grouping
p Rp Group Range Significant
10 3.66 1-10 6.67 Yes
2 3.09 1-2 2.51 No
3 3.25 1-3 3.69 Yes

Group 1 Averages: Highest Grouping
Representation Average of % Allowed p = 3; Rp = 3.25 Range = 3.69
Vietnam Veterans of America 27.33


Paralyzed Veterans of America 24.82

The first table is a calculation of ranges from Representation, a comparison of

R
p
to the calculated, and a determining of
statistical significance. First, a comparison calculated from a range from Vietnam Veterans of America to State Service
Organizations, which gave a p of 10, since there are 10 Veterans Service Groups in the group, and a range of 6.67. Also,
6.67 is greater than 3.66, so the calculated range is statistically significant. In other words, the average of State Service
Organizations statistically differs from that of Vietnam Veterans of America. The method was repeated, which can be
seen in the above table.
To create a grouping, I started at a logical point, calculated ranges, compared to

R
p
, determined if it was statistically
significant, and, if No, I grouped the Representative. As can see from the above tables, I continually compared different
Veterans Service Organizations with Vietnam Veterans of America, 1, to see which were or were not statistically
significant. In the case of Vietnam Veterans of America, I reached a statistically significant value when I compared Vietnam
Veterans of America with Military Order of Purple Heart, 3. As such, Military Order of Purple Heart, and all other
representatives after Military Order of Purple Heart, were not included in the Highest group with Vietnam Veterans of
America. Therefore, only Paralyzed Veterans of America was included. Finally, all the Veterans Service Organizations seen
in the above table: Group 1 Averages: Highest Grouping could not be separated statistically. When considering average
percent allowed, Paralyzed Veterans of America is the same as Vietnam Veterans Of America when considering
Duncan multiple-range test.
Now, I provide the other groups. After the groups, I provide the equations used.

Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 9 of 11
Group 2 Averages: Higher Grouping
p Rp Group Range Significant
10 3.66 2-11 4.43 Yes
7 3.55 2-8 3.36 No
8 3.6 2-9 3.94 Yes

Group 2 Averages: Higher Grouping
Representation Average of % Allowed p = 8; Rp = 3.6 Range = 3.94
Paralyzed Veterans of America 24.82


Military Order of the Purple Heart 23.64
AMVETS 23.49
Disabled American Veterans 23.08
Other 22.94
Attorney 21.83
American Legion 21.46

Group 3 Averages: Intermediate Grouping
p Rp Group Range Significant
5 3.44 8-12 4.56169261790745 Yes
4 3.36 8-11 1.0762113767962 no

Group 3 Averages: Intermediate Grouping
Representation Average of % Allowed p = 5; Rp = 3.44 Range = 4.56
American Legion 21.46


Veterans of Foreign War 20.88
State Service Organizations 20.65
Agent 20.39

Group 4 Averages: Lowest Grouping
p Rp Group Range Significant
2 3.09 11-12 3.49 Yes

Group 4 Averages: Lowest Grouping
Representation Average of % Allowed p = 2; Rp = 3.09 Range = 3.49
None 16.90
Equations

R
p
= s
x
r
p


s
x
=
MSE
n


r
p
is found from statistical tables[2]. MSE and n are values obtained from the ANOVA calculation. Specifically, MSE is
listed in the ANOVA table, and n was described at the beginning of the ANOVA discussion. Specifically, it is the number of
years.

Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 10 of 11

Summary

Descriptive statistics, Analysis of Variance, and Duncan multiple-range test was used to compare twelve Veterans Service
Organizations (VSO)[1]. According to the analysis, the Veterans Service Organizations of Vietnam Veterans of America and
Paralyzed Veterans of America appear to have greater success with Board of Veterans Appeals. Also, no representation,
None, is the lowest group of all and completely isolated.

If a veteran takes the time to read this paper, he or she might agree that there are methods of picking a VSO. If I would have
had this information earlier, I would have contacted the top tier group and asked if they represent 1991 Gulf War veterans.
Why? I knew that I had strong medical evidence, but mistakes happen. As such, I expected a lengthy appeal that would
eventually lead to Board of Veterans Appeals. Now I know that Vietnam Veterans of America and Paralyzed Veterans of
America have the greatest average percent allowed when considering the Board of Veterans Appeals. With that said, the
success might be due to new changes in the law that skewed the averages. Also, those in Paralyzed Veterans of America
might have a medical condition that is easily proven to be service connected. In my opinion, the graphs[9] also show that
Vietnam Veterans of America have been consistently successful.

I hope veterans find this paper useful. If I made an error, please let me know.

Chris Harding,
100% T&P Disabled 1991 Gulf War Veteran

References:
[1] US Department of Veterans Affairs. Directory of Veterans Service Organizations. va.gov[online]. 2013. Available from:
http://www1.va.gov/vso/
[2] Johnson, Richard A.(1994) Miller & Freund's Probability & Statistics for Engineers. New Jersey: Prentice-Hall, Inc.
[3] Statsoft. Electronic Statistics Textbook. statsoft.com[online]. 2013. Available from: http://www.statsoft.com/textbook/
[4] Calvin. Math 243--Normal Quantile Plots. calvin.edu[online]. 2013. Available from:
http://www.calvin.edu/~rpruim/courses/m243/F03/handouts/normquant.pdf
[5] UWlax. Comparing multiple population means using Analysis of Variance (ANOVA) Why? ANOVA is a significance
test used to compare 2 or more unknown population or treatment means. uwlax.edu[online]. 2013. Available from:
http://www.uwlax.edu/faculty/baggett/Math_145/HANDOUTS/anova.pdf
[6] Bewick, Vic; Cheek, Liz; Ball, Jonathan. Statistics review 9: One-way analysis of variance, March, 2004. Crit Care[online].
2004. vol.8(20. pp.130-136. Available from: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC420045/ doi: 10.1186/cc2836
[7] US Department of Veterans Affairs. The Board of Veterans Appeals Annual Reports to Congress 1991-2012.
bva.va.gov[online]. 2013. Available from: http://www.bva.va.gov/Chairman_Annual_Rpts.asp
[8] StatSoft. F Table for alpha = .05. statsoft.com[online]. 2013. Available from:
http://www.statsoft.com/textbook/distribution-tables/#f05
Chris Harding
Statistical Analysis: Representation Success at Board of Veterans Appeal 11 of 11
[9] Microsoft Excel file can be E-mailed upon request. Sadly, SkyDrive Microsoft Excel has too many bugs.

Das könnte Ihnen auch gefallen