Sunteți pe pagina 1din 4

Application of Statistical Concepts in the Determination of

Weight Variation in Samples


Nathalie D. Dagmang

Institute of Chemistry, University of the Philippines, Diliman, Quezon City, 1101 Philippines

Department of Food Sciences, College of Home Economics, University of the Philippines, Diliman, Quezon City 1101 Philippines

ABSTRACT

The two main objectives of the experiment are to (1) gain an understanding of some concepts of statistical analysis
and (2) apply statistical concepts in analytical chemistry.
The results obtained from the

Introduction and more precise when it is smaller. The following is


It is important that the data used in analytical the formula of the standard deviation (s)[4]:
chemistry are reliable to be able to get correct results
in experiments and problem solving. To be able to ∑𝑛
𝑖−1(𝑋𝑖 −𝑋)
2
𝑠= √ [3]
achieve this, replicate measurements are gathered. 𝑛−1
The reliability of these measurements to be used is
then evaluated using statistical concepts. Because the experiment was done by nine
Because it is impossible to experimentally groups in the class, all of which used the same
get the mean of all the values of the population, the sources of indeterminate error (i.e., the same type of
sample mean is already the most valid estimate of the measurement but different samples) the standard
true value that can be used in the experiment. It is deviations of the nine samples are pooled to get a
one of the commonly used measures of central more accurate standard deviation of the analysis:
tendency, or the probable location of the center of the
set of values and has a value intermediate between ∑𝑛𝑖=1
1 𝑛2
(𝑥𝑖 − 𝑥̅1 )2 + ∑𝑗=1
𝑛3
(𝑥𝑗 − 𝑥̅2 )2 + ∑𝑘=1(𝑥𝑘 − 𝑥̅3 )2
𝑠𝑝𝑜𝑜𝑙𝑒𝑑 = √
the extreme members of the set. This can be 𝑛1 + 𝑛2 + 𝑛3 + ⋯ 𝑛𝑠
obtained by dividing the sum of replicate
measurements by the number of measurements in
the set[4]:
where n1 is the number of measurements in set 1, n2
is the number of measurements in set 2, and so forth.
∑𝑛
𝑖=1 Xi
The ns is the number of data sets used.
X= [2]
n
Unlike the standard deviation which only
Another statistical concept used in this measures the variability of the true value, the
experiment is the standard deviation which is the confidence limits identify the values of the ends of the
measure of variation or the degree of confidence interval, a range where the true value lies
spread/dispersion of the data gathered around the at a certain level of probability (confidence level). It
sample mean. It describes the precision of the data, can also measure the precision of the data gathered.
implying that the data is less precise when it is larger The wider the range, the less precise the data is. The

1
confidence limits can be calculated from the following values, should be tested using the Q-test which will
formula [4]: say if these should be rejected or accepted for use in
future statistical calculations. There are tabulated
𝐶𝐿 = 𝑋 ±
𝑡𝑠
[5] critical values for different confidence levels which the
√𝑛
calculated value from the Q-test should not exceed. In
the experiment, the suspect values (as seen from
where t is the tabulated value for (n-1) for a certain table 3) did not exceed the Qtab so these values were
level of probability. still used in the following calculations and the sample
mean.
The purpose of this experiment is to be able
to apply the said statistical concepts in analytical In this experiment, it is calculated that the
chemistry and to learn how to get more reliable mean of Data Set 1 is 3.6765 and its standard
measurements for future experiments. deviation is 0.1. The mean of Data Set 2, on the other
It is calculated that the sample mean (for hand, is 1.6747 and its standard deviation is 0.1. In
group 8’s data only) is 3.6747, the standard deviation addition, because it is necessary that the most
is 0.1 and the confidence limits are 3.6747 ± 0.08. accurate measurement of precision is calculated, the
The pooled standard deviation (where all data group also determined the mean and pooled standard
gathered from all nine groups were considered) is s= deviation of all the data gathered by all nine groups.
0.1 while the mean is 3.6233. The mean of these data is 3.6233 while the pooled
standard deviation is 0.1. This means that if the
observations follow the normal (or Gaussian) law of
Experimental Detail errors, 68% of the members of the population may
vary between (𝑋̅ ± s), 95% may vary between (𝑋̅±
Ten samples of 25-centavo coins are used in value
2s), and 99% may vary between (𝑋̅± 3s).[1] This is
the experiment. Using the weighing by difference
suggested by the graph of Normal distribution or the
technique, the samples are weighed in the analytical
Gaussian curve (Figure 1) which is characterized by
balance. This is done by placing the ten coins with its
the mean (the maximum of the graph and about
container (watch glass) inside the balance then
which the graph is always symmetrical) and the
pressing the on/tare button. The coins are then
standard deviation (determines the amount of
removed from the equipment one by one, recording
dispersion away from the mean). A steeper curve
the absolute value of the numbers shown by the
describes a more dispersed and less precise data set
balance then pressing the on/tare button after each
while a flatter curve shows a data set with values
removal. To avoid the transfer of moisture from the
closer to each other which is more precise.
hands to the coins, which could also contribute to the
weight measured, forceps are used to remove each
sample.
frequency

Results and Discussion

It is possible that there are members of the


data set that can be considered invalid like the
outliers, the values that differ significantly from the
other results.[3] So before continuing with the analysis, values
these suspect values, usually the highest and lowest

2
Figure 1. The Gaussian Curve[4] use the pooled standard deviation for it is more
precise because it considers the results of all nine
groups. Calculations show that the pooled standard
In this case, Data Set 1, with less replicate deviation is 0.1 while the mean is 3.6198.The
measurements (six measurements), has a wider difference in results may be due to a number of
confidence interval of 3.5765 to 3.7765 thus less factors, one of these is the different levels of moisture
precise than Data Set 2 with a confidence interval of of the 25 centavo coins which can contribute an
3.5947 to 3.7547. amount of weight. Another is the amount of heat
coming from the people surrounding the analytical
The actual weight of a 25-centavo coin balance that can slightly alter the weight measured
manufactured in 2004 is 3.6 grams. Hence, the result from the equipment. Other factors like air and dirt that
from Data Set 2 is closer to the true value and it is have gone with the coin as it is placed inside the
proven that the set with the larger n is more precise balance may also have affected the results. For this
and also more accurate. reason, it is recommended that the coins be cleaned
and handled properly before weighing. This could
also be because the coins have different dates of
Conclusions manufacture. 25-centavo coins manufactured from
1995 to 2004 are made of pure brass and weigh 3.8 g
The experiment had successfully achieved while those made from 2004 until present are only
its objectives: to understand better certain statistical brass-plated steel and only weighs 3.6 g. The sample
concepts and apply these in analytical chemistry. The may be a mix of 3.8-gram and 3.6-gram 25-centavo
concepts, namely the mean, standard deviation and coins and this may greatly affect the variability of the
confidence limits, are used to determine the results. To avoid this, it is highly recommended that
approximation of the true value, the variability of the the coins be of the same manufacturing date.
data gathered and the range where the true value
likely lies.

Data set 2, which has a larger number of References


replicate measurements, has more precise results so [1] Dean, R.B. and Dixon, W.J. Simplified Statistics
its results are more reliable than that of Data set 1.
for Small Numbers of Observations. University of
Therefore, it is better to assume that the true value of
Oregon, Eugene, Ore.
the weight of a 25 centavo coin is approximately
3.6747. The variability of the data gathered, as [2] Ramachandran and Tsokos., Mathematical
measured from the standard deviation, is 0.1 while Statistics with Applications, 2009, 26-33
there is a 95% chance that the true value lies
between 3.5947 and 3.7547. It can be concluded from [3] Skoog, et al., Fundamentals of Analytical
the experiment that the larger the n (number of Chemistry, Eighth edition, 2004, 91-147
replicate measurements), the more precise and
[4] The Britannica Guide to Statistics and Probability,
reliable the results are. Hence, it is suggested that
First Edition, 305-310, 318.
there be more replicate measurements to increase
the precision and reliability of the results. [5] http://stats4students.com/measures-of-spread-
3.php. Retrieved November 19, 2010
In addition, the results of the different groups
slightly differed from each other; some had higher [6] http://sportsci.org/resource/stats/generalize.html.
means while some had lower ones. Thus, it is best to Retrieved November 19, 2010

3
[7] http://philmoney.blogspot.com/2006/12/25-
centavo-coin-new-bsp-series.html November 21,
2010

S-ar putea să vă placă și