Posted by emerjoytvale on August 12th, 2010

THE STUDY OF STATISTICS

WHY STATISTICS?

?        important in empirical studies

?        aids in decision making

?        helps to forecast or predictfuture outcomes

?        estimates unknown values

?        aids in making inferences,comparison or establishing relationship

?        summarizes data

DEFINITIONS

?        (plural sense) ? a set ofnumerical data or observations

Ex.

o  Vital statistics in a beautypageant

o  Yearly income

o  Monthly expenses

?        (singular sense) ? a branch ofscience which deals with the collection, presentation, analysisand interpretationof data.

MAIN DIVISION OFSTATISTICS

?        DescriptiveStatistics ? pertains tothe methods dealing with the collection, organization and analysis of a set ofdata without making conclusions, predictions or inferences about a larger set.

o  the main goal is simply toprovide a description of a particular data set

o  the conclusions or theimportant characteristics apply only to the data on hand

?        Inferential Statistic ? pertains to the methods dealing withmaking inferences, estimation or prediction about a larger set of data usingthe information gathered from a subset of this larger set

o  the main goal is not merely toprovide a description of a particular data set but also to make prediction and inferencesbased on the available information gathered

o  the conclusions or theimportant characteristics apply to a larger set from which the data on hand isonly a subset

Collection of Data ? refers to the method of data gathering

Presentationof Data ? refers to the process of organizing data suchas tabulation, presenting through the use of charts, graph or paragraphs

Analysisof Data ? refers to the methods of obtaining necessary,relevant and noteworthy information from the given data

Interpretationof Data ? refers to the tools of drawing of conclusionsor inferences from the analyzed data

Examples:

Descriptive Statistics

?        A college dean wants todetermine the average semestral enrolment in the past 5 school years.

?        An instructor wants to know theexact number of students who pass in his subject.

?        A school president wants todetermine the number of student-dropouts for this current school year

InferentialStatistics

?        A college dean wants toforecast the average semestral enrolment based on the enrolment for the last 5school years

?        An instructor would like topredict the number of students who will pass in his subject based on the numberof failures last year

?        A school president would liketo estimate the number of student dropouts next school year based on thecurrent data.

DESCRIPTIVE STATISTICS

POPULATION AND SAMPLE

?        Population ? collection of allcases in which the researcher is interested in a statistical study

-         the entity that the researcherwished to understand

Examples:

?        All subjects of University ofBohol

?        All department heads in UB

?        Sample ? a portion or a subset ofthe population from which the information is gathered

-         a representation of thepopulation

Examples:

?        All students of University ofBohol  coming from the rural areas

?        All department heads in UB whohave finished Ph.D. Degree

?        Parameter ? a numerical characteristicof a population

-         denoted by small Greek letters

Examples:

?        ? - population mean

?        σ ? population standard deviation

?        Statistic ? a numericalcharacteristic of a sample

-         Denoted by lower case lettersof the English alphabet

Examples:

?        X ? sample mean

?        SD ? sample standard deviation

TYPESOF DATA

?        Variable ? a characteristic orattribute of persons or objects which assumes different values or label

?        Measurement ? process ofassigning the value or label of a particular variable for a particularexperimental unit

?        Experimental unit ? the person orthe object on which a variable is measured

?        Classification of Variables

? Qualitative Variable ? yieldscategorical or qualitative responses

Examples:       Civil Status(Single, Married, Widow, etc.)

ReligiousAffiliation (Catholic, Protestant, etc.)

? Quantitative Variable ? yieldsnumerical responses representing an amount or quantity

Examples:       Height,Weight, no. of children

Typesof Quantitative Variable

? Discrete Variable ? assumesfinite or countable infinite values such as 1,2,3,?

Example:         no. ofchildren (0,1,2,3,?)

no. ofstudent-dropouts

? Continuous Variable ? cannot takeon finite values but the values are related/associated with points on aninterval of the real line

Examples:       Height (5?4?)

Weight (130.42kilos)

Temperature(32.5?C)

Levelsof Measurement

?  NOMINAL LEVEL

-         Crudest form of measurement

-         Numbers or symbols are used forthe purpose of categorizing subjects into groups

-         The categories are mutuallyexclusive, that is being in one category automatically excludes inclusion inanother

-         The categories are exhaustive,that is all possible categories of a variable should be included

Examples:       Sex:                             1 ? Male                      0 ? Female

Faculty Tenure:           1 ? Tenured                 0 ? Non-Tenured

?  ORDINAL LEVEL

-         Improvement of nominal level

-         Order/rank the data in asomewhat ?bottom to top? or ?low to high? manner

Examples:       ClassStanding (Excellent, Good, Poor)

Teacher Evaluation     1 ? Poor

2? Fair

3? Good

4? Very Good

?  INTERVAL LEVEL

-         Possesses the properties of thenominal and ordinal levels

-         Distances between any twonumbers on the scale are known

-         Does not have a stable startingpoint (an absolute zero)

Example: Consider the IQ scores of four students

70, 140, 75 and 145

Here,we can say that the difference between 70 and 140 is the same as thedifferences between 75 and 145 but we cannot claim that the second student istwice as intelligent as the first.

?  RATIO LEVEL

-         Possesses all the properties ofthe nominal, ordinal and interval levels and in addition, this has or absolutezero point

-         We can classify it, place it inproper order

-         We can also compare magnitudes

Examples:             Age, Income, ExamScores

SUMMARY CHARACTERISTICS OF LEVELS

OF MEASUREMENT

Levels of Measurement           Classify           Order               EqualLimits    Absolute Zero

NOMINAL                             Yes                  No                   No                   No

ORDINAL                             Yes                  Yes                  No                   No

INTERVAL                            Yes                  Yes                  Yes                  No

RATIO                                    Yes                  Yes                  Yes                  Yes

?        Other Classification of Data

?  Raw Data                    -data in their original form and structure

?  Grouped Data             - data placed in tabular form

?  Primary Data              - measured and gathered by the researcher that published it

?  Secondary Data          - anyrepublication of data by another researcher or agency

METHODSOF DATA COLLECTION

?        Observation Method

-         Data can be obtained byobserving the behavior of persons or objects but only at a particular time ofoccurrence

-         The data obtained is called anobservational data

?        Experimental Method

-         Especially useful when onewants to collect data for cause and effects studies

-         There is actual humaninterference with the conditions or situations that can affect the variableunder study

-         Prevalent in scientific researches

?        Use of Existing Studies

-         CHED or DECS enrollment data

-         Census Data

?        Registration Method

-         Respondents provide thenecessary information in observance and compliance with existing laws

-         Our registration, birthregistration, student registration, voter?s registration

?        Survey Method

-         The desired information isobtained through asking questions

CommonForms of Survey Method

?  Personal InterviewMethod

-         There is a person-to-personcontact between the interviewer and the interviewee

-         Considered as one of the mosteffective methods of data collection because accurate and precise informationcan be directly obtained and verified from the respondents

-         Higher response rate

-         Can be administered to therespondents one at a time

?  Questionnaire Method

-         Considered the easiest methodof data collection

-         Utilizes an instrument which isthe questionnaire as a tool

-         Lower response rate

-         Can be administered to a largenumber of respondents simultaneously

GeneralClassification of Data Collection

?  Census or CompleteEnumeration

-         Method of gathering data fromevery unit in the population

-         Not always possible because ofmoney, time and effort

?  Survey Sampling

-         Method of gathering data fromevery unit in the selected sample

-         Reduces cost, greater speed,scope and accuracy

PROBABILITYAND NON-PROBABILITY SAMPLING

?        Probability Sampling ? samplingprocedure in which every element in the population has known non-zero chance ofbeing included in the sample

CommonMethods of Probability Sampling

?        Simple Random Sampling (SRS)

-         Sampling procedure in whichevery element in the population has an equal chance of being included in the sample

-         Select n units out of N unitsin the population

-         The selection is throughlottery or the use of the table of random numbers

?        Stratified Random Sampling

-         The population of N units intosubpopulations called strata which consists of more or less homogeneous units

-         Perform simple random from eachstrata, the selection of which is independent in different strata

-         Requires a so-called ?stratification variable?

Example: Consider a population consisting of all UB students

?  Stratify the population by colleges (stratification variable)

 Commerce
 Education
 Liberal Arts
 Nursing
 Eng?g

?  Stratify the population by year level (stratificationvariable).

 I
 V
 IV
 III
 II

?        Systematic Sampling

-         Done with a random start

-         Select the sample by takingevery kth unit from an ordered population

-         k is called the samplinginterval and 1/k is the sampling fraction

Example: Suppose we select n=12 students from a population of N=50.

To employ systematicsampling, divide N by n to get k, that is

K = N/n = 50/12 = 4.667 ≈ 5

 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50

We chose every 5th unit. Thus if the random start r = 9thunit, then the sample comprise of students number

9,         14,       19,       24,       29,       34,       39,       44,       49 and 4.

?        Cluster Sampling

-         A variation of stratified samplingwhere the strata correspond to clusters, however with heterogeneouscharacteristics or attributes.

Example: A college unit may be considered as one cluster.

?        Multi-stage Sampling

-         A sampling procedure whereinthe population is divided into a sequence of sampling units corresponding tothe different sampling stages.

Common Methods of Non-Probability Sampling

?        Purposive Sampling

?        Quota Sampling

?        Convenience Sampling

PRESENTATION OF DATA

Data can be represented in variousmodes such as the textual, tabular and graphical displays.

?        Textual

-         Here, the data are presented byuse of texts, phrases or paragraphs

-         Very common among newspaperstories, depicting only the salient findings

?        Tabular

-         This is a more reliable andeffective way of showing relationships or comparisons of data through the useof tables

-         In many cases, the tables areaccompanied by a short narrative explanation to make the facts clearer and moreunderstandable

?        Graphical

-         Most effective way ofpresenting statistical findings by use of statistical graphs

-         Sets the tone for clearerinterpretation of findings especially in making comparisons

CommonForm of Graphs

?        Bar Graph ? uses rectangular barsthe length of which represents the quantity or frequency for each type category

Example: We are given the ff. enrolment data of XYZ College for the AcademicYear 1995-1996

 Year Level Males Females Total Percentage First Year 682 435 1117 10.17 Second Year 496 536 1032 27.88 Third Year 314 419 733 19.80 Fourth Year 435 385 820 22.15 Total 1927 1775 3702 100.00

We can draw the bar graph by placingthe year levels on the horizontal axis and the frequency on the vertical axisas follows:

e

n

r

o

l

m

e

n

t

(Year Level)

?        Multiple Bar Graph     ? useful when the researcher wants tocompare figures

? uses legend to guide the viewer inanalyzing the data

Example: For the same dataset, the multiple bar graph is shown below.

?        Pie Chart         - used to present quantities that makeup a whole

- the slices of the pie are drawn in proportion to the different values each class,item, group or category

- the total area of the pie is 100%

?        Line Chart ? especially useful inshowing trends over a period of time

Example: The following data shows the number of student-dropouts of UB from 1995-1999

 Year Number of Student-Dropouts (Male) Number of Student-Dropouts (Female) 1995 40 26 1996 38 43 1997 53 55 1998 48 32 1999 62 49

Theline chart of the above data set is shown below:

THE FREQUENCY DISTRIBUTIONTABLE

Datain its original form and structure are called raw data.

Example: Consider the following final exam scores of 40 students

 79 83 84 62 62 43 72 48 46 59 93 64 59 32 54 45 55 45 76 72 40 51 72 83 49 62 85 74 40 74 49 65 38 85 77 63 38 43 63 69

When these scores are arrangedeither in ascending or descending magnitude, then such an arrangement is calledan array. It isusually helpful to put the raw data in an array because it is easy to identifythe extreme values or the values where the scores most cluster.

When these data are placed into asystem wherein they are organized, then these partake the nature of groupeddata. This procedure of organizing data into groups is called a frequency distribution table (FDT).

Example: The following presents a frequency distribution table of the scoresof seventy-five Statistics students.

 Scores Frequency 10-19 5 20-29 14 30-39 23 40-49 22 50-59 11 75

Some Basic Terms

?        Class Interval              - the numbers defining the class

- consists of the end numberscalled the class limits namely the upper limit and the lower limit

?        Class Frequency          - shows the number of observationfalling in the class

?        Class Boundaries        - these are true class limits

-LOWER CLASS BOUNDARY (LCB) is defined as the middle value of the lower classlimits of the class and the upper class boundary of the preceding class

-UPPER CLASS BOUNDARY (UCB) is the middle value between the upper class limitsof the class and the lower limit of the next class

?        Class size                     - the difference between the upper limits of theclass and the preceding class

?        Class Mark                  - midpoint of a class interval

?        Open-ended class        - one which has no lower limit or upperlimit

?        Cumulative frequency - shows the accumulated frequencies ofsuccessive classes

- GREATERTHAN CF (

- LESSTHAN CF (

Example:

 Class Interval Freq LCB UCB CM >CF 10-19 5 9.5 19.5 14.5 5 75 20-29 14 19.5 29.5 24.5 19 70 30-39 23 29.5 39.5 34.5 42 56 40-49 22 39.5 49.5 44.5 64 33 50-59 11 49.5 59.5 54.5 75 11

Steps in Constructing aFrequency Distribution Table (FDT)

Step 1: Determine the number of classes. For first approximation, itis suggested to use the STURGES APPROXIMATION FORMULA.

K= 1 + 3.322 log n

where

K= approximate number of classes

n= number of cases

Step 2: Determine the ranges R.

R= maximum value ? minimum value

Step 3: Determine the approximate size, i using the formula.

i = R/K

It isusually convenient to round off i to a convenient number.

Step 4: Determine the lowest class limit (or the first class). Thisclass should include the minimum value in the data set.

Step 5: Determine all class limits by adding the class size i tothe limits of the previous class.

Step 6: Tally the scores/observations falling in each class.

Example: Constructing the FDT of the final exam scores of 40 students.

Step 1:             K        =           1+ 3.322 log n

=           1 + 3.322 log (40)

=           1 + 3.322 (1.60205)

=           1 + 5.32204

=           6.3220

K       =           6

Step 2:             R        =           max? min

=           93 ? 32

R        =           61

Step 3:             i          =           R/K

=           61/6

=           10.167

i          =           10

Step 4: Let us decide to start at the minimum value. Thus, thelowest class is the class 32 ? 41

Step 5: The classes are constructed by adding 10 to each class limit

32 ? 41

42 ? 51

52 ? 61

62 ? 71

72 ? 81

82 ? 91

92 ? 101

Step 6: Determine now the frequency of each class by tallying thescores falling in each class.

 Classes Tally Frequency 32 ? 41 | | | | 5 42 ? 51 | | | |   | | | | 9 52 ? 61 | | | | 4 62 ? 71 | | | |   | | | 8 72 ? 81 | | | |   | | | 8 82 ? 91 | | | | 5 92 ? 101 | 1 n = 40

We now proceed constructing the complete frequency distributiontable.

 Classes Frequency LCB UCB CM >CF 32 ? 41 5 31.5 41.5 36.5 5 40 42 ? 51 9 41.5 51.5 46.5 14 35 52 ? 61 4 51.5 61.5 56.5 18 26 62 ? 71 8 61.5 71.5 66.5 26 22 72 ? 81 8 71.5 81.5 76.5 34 14 82 ? 91 5 81.5 91.5 86.5 39 6 92 ? 101 1 91.5 101.5 96.5 40 1

Graphs Associated withthe Frequency Distribution Table

?                    Histogram ? a bar graph inwhich the class boundaries are plotted (on the horizontal axis) against theclass frequencies (on the vertical axis)

The followingdepicts the histogram of the frequency distribution table of the final exam scoresof 40 students.

Frequency

Class Boundaries

?                    Frequency Polygon     ? a line chart that is constructed byplotting the class marks against the class frequencies.

? The graph is obtained by connecting the consecutive points by useof straight lines.

? The polygon is closed by adding additional classmarks at each endwith a frequency of zero.

Frequency

Class Marks

(Scores)

?                    Ogives ? graphs associated withcumulative frequencies

Cumulative Frequency

Class Boundaries

(Class Scores)

?                    Ogive ? graph where the > CFis plotted against the LCB

Cumulative Frequency

ClassBoundaries

(Class Scores)

MEASURES OF CENTRALTENDENCY

Measures of CentralTendency

?                    Show the centrality of the data

?                    Measures of the average

?                    Common measures are the mean,the median and the mode

THE MEAN

?  For Ungrouped Data

?                 Population Mean (?)

_________

N

?                 Sample Mean ( X )

_________

n

Example: The following are scores of 10 sample students:

43,      32,       72,       31,       28

25,      45,       38,       42,       38

The sample mean X is computed as:

43 + 32 + 72 + 31 + 28 + 25 + 45 +38 + 42 + 38                  394

X =----------------------------------------------------------------         =          -----      =          39.4

10                                                        10

? Approximate the mean forGrouped Data

_________

n

Where:             X         =          sample mean

fi             =          frequency of the ith class

xi         =          midpointof the ith class

n          =          numberof cases

Example:       Consider theFDT scores of 75 Statistics students

 Classes Frequency 10 ? 19 5 20 ? 29 14 30 ? 39 33 40 ? 49 22 50 ? 59 11

Then to compute forX, we construct the following table

 Classes frequency (fi) Classmark (xi) fixi 10 ? 19 5 14.5 72.5 20 ? 29 14 24.5 343 30 ? 39 23 34.5 793.5 40 ? 49 22 44.5 979 50 ? 59 11 54.5 599.5 n = 75 ∑fixi = 2787.5

Thus,

_________

n

= 2787.5 / 75

Properties of the Mean

?        The mean is the most widelyused measure, which applies only to interval/ratio data.

?        It is affected by extremevalues.

?        Since it is a calculatedaverage, and its value is determined in every observation, then the mean maynot be the actual value or number in the data set.

?        The sum of the deviations aboutthe mean is zero.

?        If a constant k is added (orsubtracted) to every observation in the data set, then the mean of the new dataset increases (or decreases) by the same constant k.

?        If a constant k is multipliedto every observation in the data set, the mean of the new data set is aconstant multiple of the original mean.

THE MEDIAN

? For Ungrouped Data

-         Denoted by Me

-         The middle most value when theobservation are arranged either in ascending or descending order

-         If a data set has an evennumber of observations, then the median is the average (the Mean) of the twomost middle values.

Examples:

a.)   Consider the ff. set of scores

12, 34, 45, 72, 38,49, 65

Putting the scores in an array;

12, 34, 38, 45, 49,65, 72

Observe that Me = 45

b.)   Consider the next ff. set ofscore

8, 16, 24, 7, 21,17, 19, 18, 5, 26

Putting the scores in an array;

5, 7, 8, 16, 17,18, 19, 21, 24, 26

Observe that the two middle most values are 17 and 18. Thus, themedian is

17+18     35

Me = ---------- = ---- = 17.5

2           2

? Approximating the median forGrouped Data

where:

LCBme             =          lower class boundary of the medianclass

c                      =          class size

n                      =          number of cases

me-1            =          less than cumulative frequency of theclass preceding the median class

fme                    =          frequency of the median class

median class    =          class where the

Example:         Consider again thefollowing FDT:

 Classes Frequency 10 ? 19 5 20 ? 29 14 30 ? 39 23 40 ? 49 22 50 ? 59 11 n = 75

To approximate the median, we construct first the

 Classes Frequency 10 ? 19 5 5 20 ? 29 14 19 30 ? 39 23 42 Median Class 40 ? 49 22 64 50 ? 59 11 75 n=75

Then we locate nextthe median class. As defined, the median class is the class where the

Thus,

LCBme            =           29.5

i                      =           10

n/2                  =           37.5

me-1              =           19

fme                        =           23

Substituting these values to the formula, we obtain:

Properties of the Median

• The median is an ordinal and positional measure
• Not affected by extreme values compared to the mean

THE MODE

?  For Ungrouped Data

?        Denoted by Mo

?        Value in the data set has thehighest frequency

Example: Consider the ff. data set whose values are IQ scores

DataSet A      :           77, 83, 91, 85, 83, 100

DataSet B      :           88, 92, 71, 88, 71, 36

DataSet C      :           96, 43, 79, 68, 83, 110

Notice that themode for Data Set A is 83; the modes for Data Set B are 88 and 71 and we saythat this data set is bi-modal. Data Set C does not have a mode.

?  Approximating the Mode for Grouped Data

a.

b.

where:

LCBme             =          lower class boundary of the medianclass

i                       =          class size

fme                    =          frequency of the modal class

fi                      =          frequency of the class preceding themodal class

f2                            =          frequency of the class following themodal class

Example: Consider again the following FDT:

 Classes Frequency 10 ? 19 5 20 ? 29 14 30 ? 39 23 Modal Class 40 ? 49 22 50 ? 59 11 n=75

To approximate themode, let us first locate the modal class, the class which has the highestfrequency. Thus, 30 ? 39 is the modal class and

LCBme             =          29.5

i                       =          10

fme                    =          23

fi                      =          14

f2                            =          22

Substitutingthe values we obtain:

1. Crude Mode ? rough

a.

b.

1. Refined Mode

For larger cases, n ≥ 100

a.      Mo = 3Me - 2

b.     Mo =  - 3 (  - Me)

OtherKinds of Mean

I.                  Weighted Mean (WM) = ∑xw / N

II.               Geometric Mean, (GM)

?        Used to derived the average ofindexes, relative values and percentages

?        nth root of theproduct of n number of values

?        antilog of the logarithms ofthe middle values multiplied by the frequencies divided by the total frequencydistribution

GM            =

=(X1 ∙ X2 ∙ X3 ∙ ∙ ∙Xn)1/n

GM            =

III.            Harmonic Mean, (HM)

?        Used for spatial measurements,lengths, areas and volumes.

?        The reciprocal of thearithmetic mean of the reciprocals of the values.

HM                  =

HM                  =

Example: 10, 15, 8, 10, 13, 12, 10, 14

Solve the following:

1.     GM            = (10 ∙ 5 ∙ 8 ∙ 10 ∙ 13 ∙ 12 ∙ 10 ∙ 14)1/8

=(262080000)1/8

=11.28

GM            =

2.     HM            =

=

=

=11.06

MEASURES OF DISPERSION ORVARIABILITY

?        Measure of Dispersion ? indicate how thedata are dispersed or scattered about the average

?        Classifications:

o  Measures of Absolute Dispersion

-         Expressed in the original unitsof the original observations

THE RANGE

-         Difference between the largestand the smallest values

-         Maximum value minus minimumvalue

-         For grouped data, the range isdefined as the difference between the upper class limit of the highest classand the lower class limit of the lowest class

-         Simplest/roughest measure

Example:                     Consider the FDT

 Classes Frequency 10 ? 19 5 20 ? 29 14 30 ? 39 23 40 ? 49 22 50 ? 59 11 n = 75

The Range                   R =59 ? 10 = 49         (for discrete)

R= 59.5 ? 9.5 = 50     (for continuous)

Properties of the Range

?        It is a weak measure because ittells only the extreme values and does not provide information on the valuesbetween

?        It is greatly affected byoutliers

?        For open-ended frequencydistributions, the range cannot be computed

THE STANDARD DEVIATION

?  For Population

?  For Sample

Example: Consider the following scores of 5 students taken as samples:

8,        6,         3,         4,         4

Then  = 8 + 6 + 3 + 4 + 4 = 25 = 5

5                  5

Thus,

 x1 x1 (x1 2 8 5 3 9 6 5 1 1 3 5 -2 4 4 5 -1 1 4 5 -1 1 ∑(x1 2 = 6

The Standard deviation is therefore:

S =

S =

S =

S =

Computational Formula for the Sample Standard Deviation forUngrouped Data

Example: To verify the computed standard deviation in the previous sample,we have

 x1 xi2 8 64 6 36 3 9 4 16 4 16 ∑x1 = 25 ∑x12 = 141

Thus,

s = 2

? Approximating the StandardDeviation for Grouped Data

where

f1= frequency of the ith class

x1= classmark of the ith class

x = mean ofthe frequency distribution

n = numberof cases

Example: Consider again the FDT:

 Classes Frequency 10 ? 19 5 20 ? 29 14 30 ? 39 33 40 ? 49 22 50 ? 59 11 n = 75

Themean of this FDT was computed to be x = 37.17. The standard deviation iscomputed as follows:

 Classes fi xi (x1- ) (x1- )2 f1(x1- )2 10 ? 19 5 14.5 37.17 -22.67 513.9289 2569.6445 20 ? 29 14 24.5 37.17 -12.67 160.5289 2247.4046 30 ? 39 23 34.5 37.17 -2.67 7.1289 163.9647 40 ? 49 22 44.5 37.17 7.33 53.7289 1182.0358 50 ? 59 11 54.5 37.17 17.33 300.3289 3303.6179

∑f1(x1- ) 2 = 9466.6675

Hence,

? Computational Formula forApproximating the Standard Deviation for Grouped Data

Properties of StandardDeviation

?        Since it is a function of themean, the standard deviation is affected by every value of the data set. Thus,it is sensitive against the presence of few extreme values.

?        If each observation of a dataset is added or subtracted by the same amount k, then the standard deviation ofthe new data set is the same as the standard deviation of the original dataset.

?        If each of the data set ismultiplied by a constant k, then the standard deviation of the new data set isequal to k times the standard deviation of the standard deviation of theoriginal data set.

OTHER MEASURES OFDISPERSION

?        Coefficient of Variation (CV)

-         Ratio of the standard deviationto its mean

-         Especially useful when onecompares the variability of one data set with another data set having differentunits

CV = s/  x 100%

?        Index of Qualitative Variation(IQV)

-         Dispersion measures forqualitative nominal or ordinal variable

-         If all the values f a variableare in one category, then there is no variation and IQV = 0

-         If all the values aredistributed evenly across the categories, IQV = 1, maximum value

IQV = k(N2 - ∑f2) / N2 (k-1)

where

k = numberof categories

N = numberof observations

MEASURES OF SKEWNESS

-         Shows the degree of asymmetryor departure from symmetry of a distribution

-         Indicates also the direction ofskewness

?  Types of Skewness

o  Positively Skewed

?  Longer tail to the right

?  More concentration of values below thanabove the mean

?  This happens when  > Me > Mo

MoMe

Example: If a teacher gives a very hard exam, then one can expect that thedistribution of scores will be positively skewed.

o  Negatively Skewed

? Longer tail to the left

? More concentration of values above thanbelow the mean

? This happens when  < Me < Mo

Me Mo

Example: A very easy exam will result to a distribution of scores which isnegatively skewed.

? Pearson?s Coefficients ofSkewness

1.      sk=x-Mox

2.      sk=3(x-Me) SD

where

x           =mean

Mo       =mode

Me       =median

SD       =standard deviation

We prefer to useFormula 2 over Formula 1 in as much as the mode does not always exist.

Example:Recall the Statistics obtained from the FDT of scores of fifteen statisticsstudents.

x           =37.17

Me       = 37.54

s           =11.31

sk=3(x-Me)s

sk=3(37.17-37.54)11.31

sk=3(-0.37)11.31

sk=-1.1111.31

sk=-0.0981

Thus, we say that thedistribution of scores is skewed to the left.

MEASURES OF KURTOSIS

?  Describes the relative flatness or peakness of a distribution

?        Platykurtic ? relatively flat

?         Leptokurtic ? usuallypeaked

?        Mesokurtic ? is between the platykurticand leptokurtic curves; approaches or look like the ?normal curve?

?  Pearson?s Coefficient of Kurtosis

Ku=x1- x 4nSD4

where

x1         =          observationsin a data set

x          =          thesample mean

n          =          no.of cases

SD       =          standarddeviation

Observe that if:

Ku < 3, then the distribution isplatykurtic

Ku = 3, then the distribution is mesokurtic

Ku > 3, then the distribution is leptokurtic

compilation of Dr. Libot

Top Searches - Trending Searches - New Articles - Top Articles - Trending Articles - Featured Articles - Top Members