Você está na página 1de 32

BUSINESS RESEARCH METHODS

MBA FALL 2014

Agenda
Inferential

Analysis

Univariate Inferential Analysis


Bivariate Inferential Analysis
Hypothesis testing
Types of Inferential tests

Inferential Analysis
Inferential

analysis is used to
generalize the results obtained
from a random sample back to
the population from which the
sample was drawn.

This

analysis is required when:

A sample is drawn by random procedure


The response rate is very high

Inferential Statistics
Inferential

statistics are the mathematical


techniques for using information from
samples to make conclusions about the
population (probabilistic statements about
population).

Measurement

from

sample

are

called

statistics.
Measurements

parameter.

from a population are called

Bivariate Inferential
Analysis
Bivariate

inferential analysis is concerned with the


relationships between two variables in the sample, that
could be expected to exist in the population from which it is
drawn.

social science researcher intends to find out the


relationship between two variables: motivation and academic
performance among Asian students at the graduate level.

The

statistical method used showed that the relationship


between motivation and academic performance is of
magnitude = 0.78

All

inferential statistics give us mathematical answers but


that leads to one basic question and that is ..

Interpreting with
Confidence
What

is the probability that the result is


real, versus the probability that it is just a
fluke?

So

what is the probability that the result


of the research study (relationship of
motivation and academic performance)
was due to random chance, not due to
any systematic research effect?

Interpreting with
Confidence
The

inferential statistics provide us with


probability of getting results like these if
the only factor at play is random chance.

This

probability (provided by inferential


statistics) is known as Statistical
Significance or Statistical Reliability
denoted by letter p.

How

the p-value is interpreted?

Interpreting with
Confidence
When

the p-value is small enough, the result


is considered to be significant or reliable. If
this value is too large then the researcher will
decide that the result is just a fluke.

In

the previous example: the relationship


between motivation and academic
performance has a magnitude of 0.78 and
the probability that the observed relationship
is due to random chance is 0.023.

How

these statistics will be interpreted?

Interpreting with
Confidence
The

researcher needs to decide how


small is enough?
What is the level of probability that the
result is a fluke researcher is willing to
tolerate?
This cutoff is called Significance level.
In behavioral sciences the significance level
is generally set at 5% (0.05).

Interpreting with
Confidence
If

the probability of coincidence


(p-value) is less than 0.05 the
result is reliable and if it is less
than 0.01 then the result is highly
reliable.

Hypothesis Testing
Research

is designed to answer a specific

question.
Science

majors score higher on tests of


intelligence than students in the general
population.

The

process of determining whether this


statement is supported by the results of
the research project is referred to as
hypothesis testing

What is a hypothesis?
A statement, claim or assertion about one or
more populations.
e.g.
The mean weight of a can of beans is 420g
50% of people take out holiday insurance
To test these hypotheses we set up a Test of
hypothesis

Types of Hypothesis
Alternate

Hypothesis

Hypothesis is a statement that expresses


relationship or differences among
variables.
It is a statement that could be accepted
or rejected.

Types of Hypothesis
Alternate hypothesis can be
Directional hypothesis or
Non-directional hypothesis
Directional

Hypothesis predicts the


direction of the difference or relationship
between two groups or two variables (or
more).

Non-directional

Hypothesis do not
predict the direction of the relationship.

5 Step Procedure for Hypothesis


Testing
Sate

the null HO and alternate hypothesis H1

Select

the level of significance .

Select

the appropriate statistical test.

Compute

the

test

statistics

(obtained

values).
Interpret

the test using significance level


and statistical significance

Hypothesis Testing

1 Set up
.
test

State
hypotheses

2 Perform test
.

Perform
Calculations

3
.

Reject H0

4
.

Conclude
Ha is true

Set
significance
level

Identify Test
Statistic

Collect
data

Make

statistical
decision

Draw a conclusion

Do Not
Reject H0

Conclude H0
may be true

How to select Inferential Tests


There

are many Statistical Tests available


to help a researcher decide if differences
we find are real or chance.

The

important factors that determine the


choice of a test are:
The type of data (nominal, ordinal, interval or
ratio).
The sample size.

How to select Significance Tests


The number of samples from which
inferences are being made one, two
or k samples (e.g. children reading
levels in New Zealand, Australia and
England 3 samples).
The amount of information we have
about
population
(mean
and
standard deviation).

How to select Significance Tests


Whether we have independent or dependent
samples. A dependent sample is one where
members of the sample are somehow related to
other members.

E.g. we want to know the amount of TV a child


watches is influenced by the amount of TV their
parents watch. We may have two samples: a
group of children and a group of parents each
child is connected to a parent in the other
group, so being a member of the sample
depends on having a relative in the other group.

Comparing Means Single


Sample
Researcher

compares the performance of


the group (sample) to the performance of
the population (assuming population data
are available) for a certain variable.

E.g.

comparing IQ scores of children who


attend after-school programs (sample) to
the IQ scores in the general population. So
hypothesis tested is whether children in
after-school program perform differently
than the children in general population.

Z - test
Z

test is a test of the null


hypothesis for a single sample
when the population standard
deviation is known and sample
size is greater
X than
30.

Z - test
Z-

test is used when the following


assumptions are held for the
sample data.

Sample size is greater than 30


Sample is normally distributed
Data is interval or ratio
Population parameters are known

Not

appropriate for small samples


(n < 30).

Example
A food manufacturer processes and cans baked
beans. The net weight of a standard can of beans
is supposed to be 420g but a random sample of
50 cans gave an average weight of 415g with a
standard deviation of 30g. Is there any evidence
that the true mean weight is not 420g? (Use a
5% significance level)
415 420
Z
4.243
1.178

How will you interpret this statistic?

T test
T

test is also a statistical test of the


null hypothesis.

test is used for small groups


(n<30) and when population
standard deviation is not know.

X
t
sX

T - test

s
sX
n
sX

Is the standard error of the sampling


distribution
s is the standard deviation for population
based on sample data.

Example

A tomato grower has developed a new variety of


tomato. The variety is supposed to give good crops
without the need for a greenhouse. One of the
supposed attributes of this tomato is that average
yield per plant is 4 kg of fruit. A gardening magazine
decides to test this claim and grows 8 plants in
controlled conditions. The yield for each plant is
carefully recorded and is as follows:

Plant 1

Yield

4.2

3.3

2.5

4.8

2.75

4.2

4.6

3.6

x 3.74

Calculate t value.

Hypothesis tests involving two means


We can have:
Two large independent samples z test
Two small independent samples T- test
Paired samples T - test

What test to use?


Sing

le
sam
ple

Lar

ge
sa

zmpl
test
e

Sm

all
sa

tmpl
test
e

Me

ans

Two

sam
Indepe ples Pair
ndent
ed
sample
sam
Lar
Sm
ples
ge s all
sa
Tw
mpl
o
e

sa
Tw
mp
o
le

Pai

red

Comparing More Than 2 Means


Analysis

of variance (ANOVA) test


the significance of the
differences between more than
two means.

That

is the difference between


the means of an outcome
variable for different categories
of the predictor variable.

Comparing Means - ANOVA


Analysis

of variance is an
inferential statistical test for
comparing the means of more
than two groups (could be used
for two groups also).

One-way

ANOVA compares the


means of groups using a single
factor (brands of coffee, types of
internet connections).

Example - ANOVA
Well-being
score

Women

Well-being
score

DE

FG

For All
individuals

Men

Mean

3.40

3.00

Mean

2.60

SS

5.20

12.00

SS

5.20

Variance

1.30

1.33

Variance

1.30

SD

1.14

1.16

SD

1.14

Example ANOVA Summary


table

Wellbeing

Between
Groups
Within
Groups

Total

Sum of Squares

df

Mean
Squar
e

1.60
n(Xw-)2+n(Xm-)2

1.60
(SS/df)

1.23
(MSB/MSw)

10.40

SSw + SSm

(n-k)

12.00

(k-1)
1.30
(SS/df)

Sig .299

Você também pode gostar