Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                

TV40303 5

Download as ppt, pdf, or txt
Download as ppt, pdf, or txt
You are on page 1of 55

SEMINAR 5

Descriptive Statistics Vs Inferential Statistics

Descriptive Statistics
Summarize raw data collected from a sample to
describe its characteristics

Inferential Statistics
Make inference/generalisation about the
characteristics of a population based on the
data collected from its sample
Descriptive Stat
INFERENTIAL
Generalisation/Inference
STATISTICS

POPULATION
Sample INFERENTIAL STAT
Descriptive Stat

Generalisation/Inference
Descriptive Statistics

Methods in summarising & presenting


sample data
• Frequency table
• Pie chart
• Bar chart
Graphical Method
• stem and leaf plots
• Histogram
• Scatter diagram

Refer to Module, p. 126 - 133


Descriptive Statistics

Methods in summarising & presenting


sample data
Measures of Central Tendency
(Mean, Mode & Median)

Numerical Method
Measures of Dispersion
(Range, Sd, Coefficient of variation)

Refer to Module, p. 135 - 142


INFERENTIAL STATISTICS

Two ways to infer from sample to population

Estimation
- To estimate the population parameters from
sample statistics
Hypothesis Testing
- To test the hypothesis that what is observed
in the sample is a reflection of the actual
phenomena in the population
Estimation – An Example
In a study, you have chosen a random sample (n = 40) and
arrived at the mean score ( x ) of a science test to be 38.2
with a standard deviation (Sd) of 8.52 .

You use this mean score to estimate the mean of the


population from which the sample is drawn.

The population mean (µ) can be estimated to lie within a


certain range, referred to as Confidence Interval, using the
following formula based on 95% Degree of Confidence:

Sd
Sd
x – 1.96 < µ < x + 1.96
√n √n
Lower limit Upper limit
Lower Limit : 38.2 – 1.96 (8.52/ √40) = 35.56
Upper Limit :38.2 – 1.96 (8.52/ √40) = 40.84

35.56 < µ < 40.84


Interpretation
At 95% degree of confidence, the population mean lies
within the interval from 35.56 to 40.84. (Confidence
Interval)

(We can be 95% sure that the interval from 35.56 to 40.84
contains the population mean score of the science test.)

Note: 99% is another commonly used degree of


confidence.
95%

35.56 40.84

Note: Proportion & Standard Deviation of a population can also be


estimated from the sample statistics.
Statistical Inference with Confidence Interval:
An Example

To study the problem-solving ability of Form 5 Science


and Arts streams students, a sample was randomly
selected from a school. The following table provides the
data obtained:

n Mean Sd Sd
Stream Score
√n
Science 65 89 4.6 .57

Arts 46 77 4.3 .63


Research Q.
Is there a significance difference in the problem-solving
ability between the F5 Sc and Arts stream students?

95% confidence interval for Sc stream students:

Lower Limit: 89 – (1.96 x .57) = 87.88


Upper Limit: 89 + (1.96 x .57) = 90.12

95% confidence interval for Arts stream students:

Lower Limit: 77 – (1.96 x .63) = 75.76


Upper Limit: 77 + (1.96 x .63) = 78.24
Graphic Presentation
Since the confidence intervals
for the Sc and Arts stream do
10 not overlap, it can be
concluded that there is a
significant difference in the
problem solving ability
9 between these two groups of
● students.
From the mean scores, we
can further conclude that the
8 Sc stream students have
higher problem-solving ability

than the Arts stream students.

Sc stream Arts stream


Hypothesis Testing:
Some Statistical Techniques

Independent –samples t-test


Paired-samples t-test
One-way ANOVA
(One-way Analysis of Variance)
Two-way ANOVA
(Two-way Analysis of Variance)
ANCOVA (Analysis of Covariance)
Pearson r
(Pearson Product-Moment Correlation)
Some General Assumptions for
Parametric Tests

• Scale of measurement: interval or ratio


• Random Sampling
• Normality of population distribution*
• Homogeneity of variances*
• Reasonably large sample size ( i.e. equal or> 30)
Note:
*Violation of the above assumptions is not expected to have major
effect if the sample size is large (i.e. the number is 30 or more) &
the comparison groups have more or less equal number of
subjects
Some statistical tests require additional specific assumptions.
Refer to some more advanced books on statistics.
Examples of Hypothesis Testing

The following examples have purposely


used very small sample sizes for ease of
understanding. In your study, you are
expected to use a bigger sample.
(1) Independent-samples t-test
Achievement Scores

You have conducted a Strategy- 20, 23, 25, 22,


research to determine the Absent 27,19,28, 24,
effect of X learning (n =10) 25, 26
strategy on achievement
when they study
academic text among
college students using
the after-only design Strategy- 29, 30, 28, 32,
(Posttest Control Group
Design). Present 33, 37, 27, 36,
(n = 10) 31, 29
You obtained the data as
shown in the table.
Independent-samples t-test

Ho : There is no significant difference between the Strategy-Absent


Group and the Strategy-Present Group in their achievement
scores
H1 : There is a significant difference between the Strategy-Absent
Group and the Strategy-Present Group group in their
achievement scores

n Mean SD

Strategy-Absent 10 23.90 2.923

Strategy-Present 10 31.20 3.327

t = - 5.213 df = 18 p = .000
0.025 0.025

Reject Ho

-5.213 o
-2.101 2.101

Rejection
region

SPSS
Report

The data descriptive in Table ______ shows that the


Strategy-Present Group (Mean = 31.20, sd = 3.33) had
obtained higher achievement scores that the Strategy-
Absent Group (Min = 23.90, sd = 2.92).

The t-test confirms that the difference between the two


groups was significant at p <.05 (t [18] = 5.213).

It is thus concluded that X learning strategy has a


positive effect on the achievement of the college
students when they study academic text.
Laporan

Data diskriptif dalam Jadual ___ menunjukkan bahawa


kumpulan strategi (Min = 31.20, sl = 3.327) telah
memperoleh skor pencapaian yang lebih tinggi
berbanding dengan kumpulan tanpa strategi (Min =
23.90, sl = 2.923).

Ujian t juga mengesahkan bahawa perbezaan ini adalah


signifikan pada p <.05 (t [18] = 5.213).

Kesimpulannya, strategi X didapati mempunyai kesan


yang positif terhadap pencapaian pelajar bila mereka
mempelajari text akademik.
(2) Paired-samples t-Test
Subject Analytic holistic

You have carried out a 1 20 29


study to determine whether 2 23 30
analytical method would be
3 25 28
different from the holistic
method in the scoring of 4 22 32
students’ essays. 5 27 33
6 19 37
You obtained the data as 7 28 27
shown in the table. 8 24 36
9 25 31
10 26 29
Paired-samples t-test
(pretest-posttest/matched pairs)

n Mean SD

Analytic 10 13.00 6.749

holistic 10 12.60 4.526

t = .429 df = 9 p = .678

Ho : There is no significant difference between the analytic method and


holistic method in scoring the students’ essays.
H1 : There is a difference between the analytic method and holistic method
in scoring the students’ essays.
0.025 0.025
Acceptance
Region

o Acceptance
-1.833 1.833 Region

Rejection
.429 region

Accept Ho
SPSS
Report

Table ____ displays the descriptive data for the


analytic method (Mean = 13.00, Sd = 6.749) and
the holistic method (Mean = 12.60, Sd = 4.526).
Mean for the analytic method was found to be
higher than the mean for the analytic method.

However, the t-test indicates that the difference


was not significant at p < .05 (t [9] = .429.

This seems to imply that the analytic method does


not differ from the holistic method in the scoring of
students’ essays.
Laporan

Jadul ____ memaparkan data diskriptif bagi Kaedah Analitik


(Min = 13.00, sl = 6.749) dan Kaedah Holistik (Min = 12.60,
sl = 4.526). Min bagi Kaedah Analitik didapati lebih tinggi
daripada min Kaedah Holistik.

Namun begitu, ujian t juga mandapati bahawa perbezaan ini


tidaklah signifikan pada aras p < .05 (t [9] = .429.

Ini bermakna bahawa kaedah analitik berkemungkinan besar


tidak berbeza daripada kaedah holistik dalam pemeriksaan
essays pelajar.
(3) One-way ANOVA (Between-Subject design)
Jadual A: Writing Scores

You have conducted a study Strategy


to determine the W X Y Z
effectiveness of 4 strategies, (n = 4) (n = 4) (n= 4) (n = 4)
W, X, Y and Z on the writing
76 43 36 37
performance of Year-5
pupils. 66 75 45 22
43 66 47 22
You obtained the data as 62 46 23 25
shown in the table.
Mean 61.70 57.50 37.75 26.50
SD 13.817 15.503 10.935 7.141

SPSS
One-way ANOVA (Between-Subject design)
Table B: Summary of ANOVA for Writing Scores

Source Sum of Squares df Mean Square F p

Strategy 3314.25 3 1104.75 7.34* .005

Error 18005.50 12 150.46

Total 5119.75 15

* Significant at p < .05

Ho: There is no significant difference in the writing performance of the


pupils due to strategy
H1: There is a significant difference in the writing performance of the pupils
due to strategy.
df = [3, 12]

.05

3.49
7.34 Reject Ho
Pos Hoc Multiple Comparison – Tukey HSD Test
Table C: Results of Tukey HSD Test on Writing Scores

Mean Difference
Strategy Strategy Strategy Strategy
Mean W X Y Z

Strategy W 61.75 4.25 24.00 35.25**


Strategy X 57.50 19.75 31.00*
Strategy Y 37.75 11.25

Strategy Z 26.50

* Significant at p <.05 ** significant at p < .001


Report

Data descriptive in Table A shows that Strategy W group (Mean


= 61.75, Sd = 13.82) obtained higher mean score compared to
Strategy X group (Mean = 57.5, Sd = 15.50), Strategy Y group
(Mean = 37.75, Sd = 10.94) and Strategy Z group (Mean = 26.50,
Sd = 7.14). Likewise, Strategy X group obtained higher mean
score than Strategy Y group and Strategy Z group, and
Kumpulan Strategi Y group also obtained higher mean score
than Strategy Z group

One-way ANOVA also confirms that there was a significant


difference in the writing performance according to strategy at
p < .05 (F [3, 12] =7.343). However, the Tukey HSD test only
reveals that there was a significant difference in the writing
performance between Strategy W group and Strategi Z group at
p <.01, and between Strategy X group and Strategy Z group at p
<.05. The other differences were not significant at p <.05.

In conclusion……..
Laporan
Data diskriptif dalam Jadual A menunjukkan bahawa Kumpulan Strategi W
(Min = 61.75, sl = 13.817) mencapai skor penulisan yang lebih tinggi
berbanding dengan Kumpulan Strategi X (Min = 57.5, sl = 15.503),
Kumpulan Strategi Y (Min = 37.75, sl = 10.935) dan Kumpulan Strategi Z
(Min = 26.50, sl = 7.141). Begitu juga Kumpulan Strategi X mendapat skor
yang lebih tinggi daripada Kumpulan Strategi Y dan Kumpulan Strategi Z,
dan Kumpulan Strategi Y juga mencapai skor yang lebih tinggi daripada
Kumpulan Strategi Z.

Ujian ANOVA satu hala juga mengesahkan bahawa terdapat perbezaan


yang signifikan antara skor penulisan mengikut strategi pada aras p < .05
(F [3, 12] =7.343). Namun begitu, hasil Ujian Tukey hanya mendapati
perbezaan yang signifikan dalam skor penulisan antara Kumpulan Strategi
W dengan Kumpulan Strategi Z pada aras kesignifikan p <.01, dan antara
Kumpulan Strategi X dengan Kumpulan Strategi Z pada aras p <.05.
Manakala perbezaan yang lain tidaklah signifikan pada aras p <.05.

Kesimpulannya, ……….
(4) 2-way ANOVA
Table A: Recall Scores
You have conducted a
Strategy-Present Strategy-Absent
study to examine the
effect of a learning Con Abs Con Abs
strategy X on the recall (n = 5) (n = 5) (n = 5) ( n = 5)
of two types of 17 12 11 12
vocabulary items 20 10 9 9
(Concret & Abstract) 14 14 12 15
16 8 8 11
17 11 13 13
You obtained the data
Mean 16.80 11.00 10.60 12.00
in the form of recall
scores as shown in the SD 2.168 2.236 2.074 2.236
table.

SPSS
Two-way ANOVA (2 x 2 Factorial Design)

Table B: Summary of ANOVA for Recall Scores by Strategy and Vocab Type

Source SS df MS F p

Strategy 33.800 1 33.800 7.116* .017


Vocab 24.200 1 24.200 5.095* .038

Strategy 64.800 1 64.800 13.642** .002


X Vocab

Error 76.000 16 4.750


Total 198.800 19

* Significant at p < .05 ** significant at p < .01


Ho: There is no significant difference in the recall scores
due to strategy
H1: There is a significant difference in the recall scores
due to strategy

Ho: There is no significant difference in the recall scores


due to vocabulary type
H1: There is a significant difference in the recall scores
due to vocabulary type

Ho: There is no significant interaction between strategy


and vocabulary type on the recall scores.
H1: There is a significant interaction effect between
strategy and vocabulary type on the recall scores
df = [1, 16]

.05

Reject Ho

4.49
13.642
5.095 7.116
Main & Interaction Effects: Meaning
Strategy- Strategy- Main Effect
Present Absent

Concrete 16.8 10.6 13.7


+ 6.2
+ 5.8 - 1.4 + 2.2
Abstract 11.0 12.0 11.5
- 1.0

Main effect 13.9 11.3


+ 2.6

There is interaction
Main & Interaction Effects: Meaning

17 Strategy-Present/Con

Strategy-Absent/Abs

Strategy-Absent/Con Strategy-Present/Abs

Concrete Abstract
Report

Table A shows………..

The two-way ANOVA in Table B confirms that there was


a significant interaction between strategy and vocabulary
typre F[1,16] =13.64, p <.01. There was also a significant
main effect for strategi, F[1,16] = 7.116 with higher mean
recall score for Strategy-present group compared to the
Strategy-absent group The main effect for vocabulary
type was also significant, F[1,16] =5.059, p <.05, with
higher mean recall score for concrete vocabulary
compared to abstract vocabulary

It is thus concluded………………
Laporan

Jadual A menunjukkan ……..

Ujian ANOVA 2-hala dalam Jadual B mengesahkan bahawa


terdapat kesan interaksi antara strategi dengan jenis
perbendaharaan kata, F[1,16] =13.64, p <.01. Terdapat juga kesan
utama yang signifikan bagi strategi, F[1,16] = 7.116 dengan skor
ingat kembali yang lebih tinggi bagi kumpulan strategi berbanding
dengan kumpulan tanpa strategi. Kesan utama bagi jenis
perbendaharaan kata juga didapati signifikan, F[1,16] =5.059, p
<.05, dengan skor ingat kembali yang lebih tinggi bagi
perbendaharaan kata konkrit berbanding dengan perbendaharaan
kata abstrak.

Kesimpulannya………..
Main & Interaction Effects: Meaning
Strategy- Strategy- Main Effect
Present Absent

Concrete 15.0 12.4 13.7


+ 2.6
+ 2.2 + 2.2 + 2.2
Abstract 12.8 10.2 11.5
+ 2.6

Main effect 13.9 11.3


+ 2.6

There is no interaction
Main & Interaction Effects: Meaning

17

Strategy-Present/Con

Strategy-Present/Abs
Strategy-Absent/Con

Strategy-Absent/Abs

Concrete Abstract
(5) ANCOVA With Strategy Without Strategy
You have conducted (n = 10) (n = 10)
a research to 57, 70, 87, 79, 93, 50, 43,
determine the effect Pretest 71, 71, 79, 57, 87, 50, 64,
of Strategy ABC on Score 50, 93, 50, 79, 57
the reasoning ability 57
of college students Mean = 68.50 Mean = 65.90
using the Pretest Sd = 14.97 Sd = 17.37
Posttest Control Posttest 63, 79, 95, 79, 95, 68, 47,
Group Design. Score 89, 79, 89, 63, 89, 58, 74,
58, 89, 68, 74, 53
You obtained the 71
data as shown in the
Mean =78.00 Mean = 70.00
table. Sd = 12.60 Sd = 15.33

SPSS
Ho : There is no significant difference between the With
Strategy Group and the Without Strategy Group in
their reasoning ability
H1 : There is a significant difference between the With
Strategy Group and the Without Strategy Group
in their reasoning ability.

Table A
Means & standard deviations of the reasoning ability pretest
and posttest scores

Pretest Posttest
n Mean Sd Mean Sd

With Strategy 10 68.50 14.97 78.00 12.60

Without Strategy 10 65.90 17.37 70.00 15.33


Table B
Adjusted means & standard errors of the reasoning ability posttest
scores

n Adjusted Mean Standard Error


With Strategy 10 76.97 1.85
Without Strategy 10 71.03 1.85
Table C
ANCOVA for the reasoning ability scores

Source SS df MS F p
Corrected Model 3281.08 2 1640.54 48.01** .000
Intercept 432.31 1 432.31 12.65* .002
Pretest 2961.08 1 2961.08 86.65** .000

Strategy 175.35 1 175.35 5.13* .037


Error 580.92 17 34.17
Total 113382.00 20
Corrected Total 3862.00 19

* Significant at p <.05 ** Significant at p < .001


Report
Table A displays the means and standard deviations of the reason ability
pretest and posttest scores for both the With Strategy and Without
Strategy groups. The pretest mean score for the With Strategy group
………….
Table B shows the adjusted means and the standard errors of the
reasoning ability posttest scores for both the With Strategy and Without
Strategy groups after including the pretest scores as the covariate. These
descriptive statistics indicate that the With Strategy group (Mean = 76.97,
SE = 1.85) outperformed the Without Strategy group (Mean = 71.03, SE =
1.85). This seems to suggest that Strategy ABC had a facilitative effect on
the reasoning ability among college students.
The above conclusion is confirmed by ANCOVA results as shown in
Table C. The results indicate that the pretest was significant at p < .001, F
[1, 17] = 86.65. This suggests that the pretest had a significant influence
on the posttest. Using the pretest as the covariate, the analysis further
indicates that there was a significant effect of Strategy ABC on the
reasoning ability, F [1, 17] = 5.13, p <.05.
(6) Pearson Product-Moment Correlation r
Subject Analytic holistic

You have carried out a 1 20 29


study to determine whether 2 23 30
there is a correlation in the
3 25 28
scoring of students’ essay
by analytical method with 4 22 32
the scoring by the holistic 5 27 33
method. 6 19 37
7 28 27
You obtained the data as 8 24 36
shown in the table.
9 25 31
10 26 29
SPSS
Correlation

Hoo: There is no significant correlation in the scoring of


students’ essays by analytic and holistic methods.
H1: There is a significant correlation in the scoring of
students’ essays by analytic and holistic methods.

df = no. of pairs – 2
r [8] = .938, p<.01
= 10 – 2
= 8
Since r [8] = .938 is greater the critical value
at p = .01 (i.e. .765), we reject Ho. There is a
high possibility that both the scoring methods Critical value
are similar to each other.
p = .05, r [8] = .632
p = .01, r [8] = .765
Correlation: Meaning

r = +.65
r = +1.00

r = - 1.00
r = -.65
(7) Chi-square X2 [Nominal Data]

You have collected data on Yes No Total


opinion whether there is too
much violence on TV from Male 15 38 53
male and female
undergraduates. You want
to know whether male and Female 30 20 50
female differ in their
opinions.
Total 45 58 103
The data are given in the
table.
CHI-SQUARE

Ho: There is no significant difference between the


opinions of male and female undergraduates
regarding violence on TV
H1: There is a significant difference between the
opinions of male and female undergraduates
regarding violence on TV

df = (r – 1)(c – 1) = (2 -1)(2-1)
=1
X2 = 10.5077 p = .001

SPSS
The Chi-square test suggests that there
is a significant difference in the opinions
between male and female
undergraduates regarding violence on
TV, X2 [1] = 10.508, p <.01, with more
females considering there is too much
df = 1 violence than males.

.01
Reject Ho

6.63
10.508
VassarStats Statistical
Computation Website

http://faculty.vassar.edu/lowry/VassarStats.html
Parametric Tests Vs Non-Parametric Tests
(Violation of assumptions & Small Sample Size)
Parametric Statistical Non-Parametric Statistical
Techniques Techniques
Independent-samples t-test Mann-Whitney test

Paired-samples t-test Wilcoxon (matched-pairs) signed-ranks


test
One-way ANOVA Kruskal-Wallis Analysis of Variance of
(Between-Subjects Design) ranks
One-way ANOVA Friedman Analysis of Variance by ranks
(Within-Subjects Design) test
Analysis of Covariance Rank Transformation, followed by
(ANCOVA) ANOVA
Pearson Product-Moment Spearman’s Rank-Order Correlation
Correlation
GOOD LUCK &
HAPPY STUDYING

THANK YOU

You might also like