Katharine Riche
Introduction………………………………………………………………………………….page 4
Methodology………………………………………………………………………………...page 5
Results/Discussion…………………………………………………………………………..page 7
Conclusions…………………………………………………………………………...……..page 8
Appendix…………………………………………………………………………………….page 9
Excel Spreadsheet………………………………………………………………..………...page 10
PROC UNIVARIATE
a. Notes…………………………………………………………………...page 11
b. Graphs………………………………………………………………….page 14
a. Notes…………………………………………………………………...page 24
b. Graphs…………………………………………………………….……page 27
a. Notes…………………………………………………………………page 37
b. Graphs………………………………………………………………..page 40
2
d. Two Sample T-tests………………………………………………….....page 47
a. Notes…………………………………………………………………...page 50
b. Graphs……………………………………………………………….…page 53
3
Introduction
Three researchers by the names of Eagle, Asa, and Garrott designed an experiment that
could potentially help manage the wild horse population on federal lands. They suggested hat
the dominant male in each group should be sterilized as a way to reduce foaling rates for two or
more years. They designated two herd management areas, one in Flanigan in northwest Nevada
and another in Beauty Butte in southeast Oregon. In December of 1985 they rounded up wild
horses in bands, counted them, determined their sex, established their age by tooth wear, and
photographed and put numbered collars on the horses that were three years of age or older.
Then, they identified the dominate male in each band and performed the vasectomy and put a
radio-transmitting collar on him. Finally, they released the horses in the designated areas. Both
Flanigan and Beauty Butte also had bands of horses whose dominant males were not sterilized.
This acted as a control so that the researchers could compare data and determine whether or not
sterilizing the males was an effective way to reduce foaling rates. For the next three years, the
researchers went to Flanigan and Beauty Butte three or four times each year. Each time, they
located the sterile males and recorded the number of adult, the number of sterile males, and the
number of foals. They also recorded the number of adults and the number of foals in bands with
no sterile males. After three years, the researchers had a set of data that included the number of
adults, the number of sterile males, the number of foals, the year the band of horses was located,
the location of the horses, the date of observation, and the treatment type: 0=not sterilized and
1=sterilized (see Excel Spreadsheet). The following report is a product of organizing the data,
forming questions, and running tests on SAS Analyst to determine if sterilizing the dominant
males was an effective way to reduce foaling rates for two or more years.
4
Methodology
To analyze the data, I first looked at the data and organized the information. During the
year of 1986, the researchers went to the Flanigan management area three times and the Beauty
Butte management area three times. Each time, they recorded the number of adults and foals for
both sterilized and unsterilized groups. In 1987, they went to Flanigan three times and Beauty
Butte four times and recorded the number of adults and foals for both sterilized and unsterilized
groups. In 1988, they visited both areas three times and recorded the number of adults and foals
for both groups. Looking at the data they recorded over the course of three years, it was clear
that in both management areas, the number of adults in sterilized groups was less than the
number of adults in unsterilized groups. This was also true for the number of foals. Also, when
comparing data recorded for each year, the number of foals just from sterilized groups decreased
Next, I ran PROC UNIVARIATE on SAS Analyst to determine whether or not the
populations of wild horses were normally distributed. I chose to look at the number of adults
within each location, the number of foals within each location, the number of adults from groups
of both treatment types, and the number of foals from groups of both treatment types. Looking
at the two locations could help the researchers see whether or not one location was a better
environment for wild horses than the other. Looking at sterilized groups of wild horses versus
unsterilized groups of wild horses could help the researchers determine whether or not the
number of adults and foals were influenced by the treatments. I went through each set of data
that I chose to look at and found that based on the information and the graphs, the variables
range from being close to normal to further away from normal. I think this range is due to the
wide differences in numbers of adults and foals. These differences seem to be due to the type of
5
treatment the group was given. There are greater than 100 adults recorded each time for
unsterilized groups and less than 100 adults recorded each time for sterilized groups (see Excel
Spreadsheet). After I looked at the number of adults and foals for each location and the number
of adults and foals for each type of treatment, I thought about how the number of sterile males
affects the number of foals, but since the type of treatment affects the number of sterile males, I
did not make that a separate set of data to look at on SAS. I also assumed that the number of
sterile males was probably incorporated into the number of adults, but that information was not
clear.
Skewness is supposed to equal 0 to be normal, but all of them had a positive skew.
Kurtosis is also supposed to equal 0, but some were positive, or leptokurtic, and others were
negative, or platykurtic. The mean, median, and mode of each set of data are supposed to be
equal to each other, but most are slightly off or more than slightly off. I made notes on each part
of the PROC UNIVARIATE and on the graphs (see PROC UNIVARIATE, notes and graphs).
Most of the graphs were not very close to being normally distributed and some were very far off.
Even though I was not sure whether or not the information was close enough to normal, I ran F
tests and T tests for all of the scenarios and came up with one-tailed and two-tailed questions
(see PROC UNIVARIATE, Two Sample F-tests and T-tests). These questions were based on
variance, mean, the influence of location on the adult and foal populations, and the influence of
treatment on the adult and foal populations. Lastly, I evaluated the data produced from running
the F and T tests and answered the question by setting up the null hypothesis and the alternate
hypothesis. By using the data I was able to write a statistical conclusion and an English
conclusion.
6
Results/Discussion
First, I looked at the influence of location on the number of adults. After looking at
variance among both locations, I concluded that the variances were the same and therefore I used
the equal variance T-Test. I asked the question: Is the mean number of adults in Beauty Butte
and Flanigan the same? My English conclusion was probably, the mean number of adults in
Beauty Butte and Flanigan are the same. This showed that although the recorded number of
adults in both locations had a wide range due to treatment type, the location itself did not have a
huge influence on the means. I also did both one-tailed T-tests to see if one location had a
greater influence on the number of adults than the other. Location is important because if the
location is not ideal for wild horses (i.e. if there is a lack of food or water), than it could have an
effect on the number of horses as a group and on the foaling rates (see Adults, Location B and
location on the number of foals. An F-test and different T-tests were completed to look at
similar questions I asked for the adults (see Foals, Location B and Location F, PROC
completed an F-test and concluded that I should use the equal variance T-test. I asked the
question: Is the mean number of adults from the sterilized group the same as the number of
adults from the unsterilized group? My English conclusion was probably, the mean number of
adults is not the same for both treatments. This conclusion showed that the type of treatment did
have a significant influence on the number of adults that were recorded during each visit to the
management areas. I did both one-tailed T-tests to see if one treatment had a greater effect on
the number of horses than the other (Adults, Treatment 0 and Treatment 1, PROC
7
UNIVARIATE, Two-Sample T-tests). I also looked at the influence of treatment type on the
number of foals and completed an F-test and T-tests (Foals, Treatment 0 and Treatment 1, PROC
Conclusions
The data gathered by Eagle, Asa, and Garrot did support their suggestion that sterilizing
the dominate males in wild horse populations would reduce the foaling rates for two or more
years. By looking at the data they recorded alone, I concluded that although location may not
have had a significant impact on the number of adults and foals, the type of treatment (sterilized
or unsterilized) did. In general, there were far less adults and foals in sterilized groups compared
to the adults and foals in unsterilized groups. This was true for both locations. As I continued to
look at the data and ask questions based on variance, mean, and whether or not location or type
of treatment had an effect on the number of adults or foals, I was able to draw up a few more
conclusions.
The data seemed to come from populations that were close to normally distributed and
possibly even non-normally distributed. The conclusions for the F-tests and T-tests about
location seemed to coincide with what I inferred from the dataset, which was that location did
not have a large impact on the number of adults and foals. The conclusions for the F-tests and T-
tests were not so easily made because there was an unequal variance, which led me to use the
unequal variance T-test and I had some undetermined results. However, this helped me realize
that there may have been a lot of room for error during this experiment.
In 1986, the researchers went to both locations three separate times. Each time they
went they counted the number of adults, the number of sterile males (may have been 0 if the
group they were counting was part of the unsterilized group), and the number of foals. Before
8
they began this though, they photographed and collared all the horses that were 3 years of age or
older and I assumed that the horses that were less than 3 years of age would be counted as foals.
They did not mention whether or not they rechecked tooth wear on previously distinguished foals
each year they visited. This is important because many of them could have turned three years
old or more during the course of this experiment, but the researchers may have been counting
them as foals and not adults. Also, since the population size of sterilized groups were smaller
from the beginning, it is hard to say whether or not the birth rate decreased by a lot because the
ratio of the numbers may be fairly close. There are also many confounding factors such as diet,
climate, and health that could have affected the populations over the three year study.
9
Appendix
*The original appendix of this paper was made up of printed out SAS Analyst
procedures that had hand written statistical analyses and conclusions. The
following is a few examples of descriptive and inferential statistics.
PROC UNIVARIATE:
The UNIVARIATE Procedure
Variable: Adults (Adults)
Location = B
Moments
N 20 Sum Weights 20
Mean 102.35 Sum Observations 2047
Std Deviation 56.7786793 Variance 3223.81842
Skewness 0.2545658 Kurtosis -1.6895546
Uncorrected SS 270763 Corrected SS 61252.55
Coeff Variation 55.4750164 Std Error Mean 12.6960987
Location Variability
Quantiles (Definition 5)
Quantile Estimate
10
0% Min 36.0
Extreme Observations
----Lowest---- ----Highest---
36 17 169 6
38 20 170 9
40 16 173 7
44 18 178 10
45 19 192 8
Moments
N 18 Sum Weights 18
Mean 119 Sum Observations 2142
Std Deviation 85.264295 Variance 7270
Skewness 0.28057448 Kurtosis -1.7282742
Uncorrected SS 378488 Corrected SS 123590
Coeff Variation 71.6506681 Std Error Mean 20.0969871
Location Variability
Quantiles (Definition 5)
Quantile Estimate
11
100% Max 243.0
99% 243.0
95% 243.0
90% 240.0
75% Q3 197.0
50% Median 100.5
25% Q1 36.0
10% 30.0
5% 25.0
1% 25.0
0% Min 25.0
Extreme Observations
----Lowest---- ----Highest---
25 32 197 25
30 36 226 24
31 34 232 21
35 33 240 28
36 37 243 27
F-TEST:
Two Sample Test for Variances of Adults within Location
Sample Statistics
Location
Group N Mean Std. Dev. Variance
--------------------------------------------------
B 20 102.35 56.779 3223.818
F 18 119 85.264 7270
Hypothesis Test
- Degrees of Freedom -
F Numer. Denom. Pr > F
----------------------------------------------
0.44 19 17 0.0896
12
TWO SAMPLE EQUAL-VARIANCE T-TEST:
Sample Statistics
Hypothesis Test
13