Basic Inferential Statistics

October 7, 2023
 Inferential Statistics
• It is about testing hypothesis.
• Establishing null and alternative hypotheses
can be a frustrating and confusing process.
• H0 = initially is assumed to be true.
– Null means that nothing is new.
– Analogous to the courtroom ideal of innocent
until proven guilty.
 Approaches to
Hypothesis Testing

• Classical/tabular
approach
• p-value approach
 p-value approach

• Alternative to classical approach

(tabular approach)
• Focus is on the probability (p-value)
that you would be wrong if you reject
the hypothesis – not on the computed
and the standard test statistic
 How is it done?

• Generate the p-value of a certain test

– usually done through statistical
softwares (e.g., SPSS, STATA, SAS)
• Compare the p-value to the level of
significance (usually 0.05)
 p-value approach

• Make a decision about the 𝐻0

– If p-value is  to 0.05; the
result of the test is
significant; reject 𝐻0
– Do not reject 𝐻0 , if
otherwise
• Connect with what others
have*
• The performance of the 3 groups significantly differ.
• Reject H0
• But, is it the graduates with significantly diff. performance compared to
the undergraduates? Or is it the post graduates against the
undergraduates?
 Testing hypotheses
• We either reject the null hypothesis or fail to
reject the null hypothesis.
– Type 1 error: an error committed when a true null
hypothesis is rejected.
– Type 2 error: an error committed by failing to
reject a false null hypothesis
• Example:
– In a manufacturing company (production line).
– In an organization (corrupt worker)...
 Inferential Statistics
• Hypothesis Testing
• Types of error
– Type 1 error – when we reject the null hypothesis when in
fact it is true.
– Type 2 error – when we accept the null hypothesis wherein
in fact it is false.
• This is the reason why we have “level of significance”
or “margin of error”.
 Level of Significance / Margin of error

Example: 1%; 5%; 10%

• 10% - meaning you are 95% confident that you have made the
right decision.. 5% chances in 100% that you commit a wrong
decision of rejecting a hypothesis wherein fact it is true.

NOTE: In social research, we allow greater amount (%) of error

because usually, the respondents answer based merely on
“memory recall”.
• Therefore, the level of significance is the maximum probability
value of making or wrong decision.
 Statistics

Descriptive Inferential Probabilistic

Test of Difference
Findings for central Parametric Non-parametric Poisson
location: Regression (rare
Z- test Chi-square
• Mean occurrence)
• Median T-test
• Mode ANOVA

Test of Relationship

Parametric Non-parametric
Measures for variability
• Range Linear-Correlation Spearman Rank
Correlation
• Mean Absolute
Deviation
• Standard Deviation Test of Association
• Variance
Parametric
Linear Regression
 What is data analysis?

Making sense out of a set of

raw observations with the aim of addressing the
goal/s of the research
 Data Analysis
• In data analysis, how can you decorrelate error?
• Answer
– Follow the “Principle of Independent Judgment”.
– Tame the halo effect.
Imagine…
 Decorrelate error.
• Some individuals greatly overestimate the true
number, others underestimate it, but when many
judgments are averaged, the average tends to be
quite accurate.
• All individuals look at the same jar, and all their
judgments have a common basis.
• On the other hand, the errors that individuals make
are independent of the errors made by others, and
they tend to average to zero.
 Decorrelate error.
• However, the magic of error reduction works well
only when observations are independent and their
errors uncorrelated.
• If the observers share a bias, the aggregation of
judgments will not reduce it.
• Allowing the observers to influence each other
effectively reduces the size of the sample.
 Decorrelate error.
• To derive the most useful information from multiple
sources of evidence, you should always try to make
these sources independent of each other.
 Decorrelate error.
• This rule is part of good police procedure.
• When there are multiple witnesses to an event,
they are not allowed to discuss it before giving their
testimony.
 Topics
• T-test (Independent and Dependent)
• Linear Correlation
• Linear Regression
• Analysis of Variance
 T-test
(dependent and independent)
 Example of T-Test (dependent)
• The data below are from participants’ ratings of how jittery they feel before and
after drinking Red Bull (higher scores more jitter). Calculate the problem using
alpha 0.05 level of significance to assess whether or not Red Bull increased feelings
of jitteriness.

Before After
3 8
2 7
4 9
2 8
3 5
6 7
9 8
8 8
 Testing of Hypothesis
1. Problem Statemen
2. What is the hypothesis?
3. Statistical tool to be used
4. Level of significance
5. Computation
6. Decision
7. Conclusion
 Example of T-Test (dependent)

• Problem: Is there are significant increase of participants’ rating of how jittery they
feel before and after drinking Red Bull?
• Hypotheses:
– Null: There is no significant increase in participants’ rating before and after…
– Alternative: There is a significant increase of participants’ rating before and after…

• Statistical Tool: T-test for dependent samples

• Level of significance: Alpha at .05
 Example of T-Test (dependent)
Before After Difference (D)²
3 8 -5 25
2 7 -5 25
4 9 -5 25
2 8 -6 36
3 5 -2 4
6 7 -1 1
9 8 1 1
8 8 0 0
Σx = 37 Σx = 60 ΣD = -23 ΣD2= 117
D = 2.875

D 2.875
t = ------------------------------------- t = ---------------------------
√[(117 – (529 /8)] / 8(8-1)
√[(𝞢D² - (𝞢D)² /n] / n(n-1)
Answer: 3.06
n-1 = 8
Alpha = .05
 Example of T-Test (dependent)
• Decision
– Since the computed value (3.01) is greater than the tabular
value (2.306), REJECT Null hypothesis.

• Conclusion
– There is a significant increase of feeling of “jittery” for
participants before and after drinking Red Bull.
 Example of T-Test (independent)
• The following are the scores of 10 males and females training attendees. Test the null
hypothesis that there is no significant difference between the performance of male and
female attendees in the said test. Use the T-Test at .05 level of significance.

Male (x1) Female (x2)

14 12
18 9
17 11
16 5
4 10
14 3
12 7
10 7
9 6
17 13
 Example of T-Test (independent)

• Problem: Is there are significant difference between the performance of male and
female attendees in the said test?
• Hypotheses:
– Null: There is no significant difference between the performance of male and female attendees in the
said test.
– Alternative: There is a significant difference between the performance of male and female attendees
in the said test.

• Statistical Tool: T-test for independent samples

• Level of significance: Alpha at .05
 Example of T-Test (independent)
(Σx1)2
Sum of Squares1 = Σx21 - --------------------
n1
(Σx2)2
SS2= Σx2 2- --------------------
n2

x1 – x2
t = -------------------------------------------------------
√(SS1 +SS2 / n1 + n2 – 2) (1/n1 + 1/n2)
 Example of T-Test (independent)
Male (x1) Female (x2) Male (x1)2 Female (x2)2
14 12 196 144
18 9 324 81
17 11 289 121
16 5 256 25
4 10 16 100
14 3 196 9
12 7 144 49
10 7 100 49
9 6 81 36
17 13 289 169
Σx1 = 131 Σx2 = 78 Σx 21 = 1891 Σx 22 = 738
n1 = 10 n2 = 10
Mean x1 = 13.1 Mean x2 = 7.8

x1 – x2
t = -------------------------------------------------------
√(SS1 +SS2 / n1 + n2 – 2) (1/n1 + 1/n2)
Answer: 2.88
n-1 = 9
Alpha = .05
 Example of T-Test (independent)
• Decision
– Since the computed value (2.88) is greater than the tabular
value (2.262), REJECT Null hypothesis.

• Conclusion
– There is a significant difference between the performance
of male and female attendees in the said test.
Application of Linear Regression
 Linear Trend Projection
• Used for forecasting linear trend line
• Assumes relationship between response
variable Y & time X is a linear function
 Correlation

• Answers ‘how strong is the linear

relationship between 2
variables?’ 𝑛 σ 𝑥𝑦 − σ 𝑥 σ 𝑦
𝑟=
2 2 2 2
• Coefficient𝑛 σof
𝑥 − (σ 𝑥) 𝑛 σ 𝑦 − (σ 𝑦)
correlation used
 Degree of Correlation

Range (+ or -) Description
.81 – 1.0 Very High
.61 - .80 High
.41 - .60 Moderate
.21 - .40 Low
.0 - .20 Very low
 Simple Linear Regression
Formulas for Calculating “a”
and “b”
a = y - bx

n xy -  x  y
b=
n x 2 - ( x ) 2

𝑛 σ 𝑥𝑦 − σ 𝑥 σ 𝑦
𝑏=
𝑛 σ 𝑥 2 − (σ 𝑥)2
 Simple Linear Regression Model:

y = b0 + b1 x

Where:
• y = the predicted value
• b0 = sample intercept (y-intercept)
• b1 = sample slope (estimator of the slope of
the population regression)
• x = the predictor value
Exercise (1)
Exercise (2)
Next Topic: Analysis of Variance
(ANOVA)
 F-Test and ANOVA (Analysis of Variance)

• Analysis of Variance – is a technique in inferential

statistics designed to test whether or not more than
two samples are significantly different.
 F-Test and ANOVA (Analysis of Variance)
• Below are the scores of the applicants during the
entrance exam in October 2021.
Student Group 1 Group 2 Group 3
1 84 70 90
2 90 75 95
3 92 90 100
4 96 80 98
5 84 75 88
6 88 75 90
 Example of T-Test (independent)

• Problem: Is there are significant difference among the performance of the 3 groups
who took the entrance examination?
• Hypotheses:
– Null: There is no significant difference among the performance of the 3 groups who took the entrance
examination
– Alternative: There is a significant difference among the performance of the 3 groups who took the
entrance examination

• Statistical Tool: One Way ANOVA

• Level of significance: Alpha at .05
SS = Sum of Squares
MS = Means of Squares
SSC = Sum of Squares Columns
SSE = Sum of Square Errors
SST = Total Sum of Squares
MSC = Treatment Variance
MSE = Error Variance
Critical Value of the F-Distribution

Answer: 13
Numerator degrees of freedom: 2
Denominator: 13

Table: 3.81
Answer: 11.37
Numerator degrees of freedom: 2
Denominator: 13

Table: 3.81
 Example of One Way ANOVA
• Decision
– Since the computed value is greater (11.31) than the
tabular value (3.81), REJECT Null hypothesis.

• Conclusion
– There is a significant difference among the performance of
job applicants in terms of examination scores.
• Thank You.