RESEARCH AND STATISTICS

Chapter 6
BEYOND DESCRIPTIVE
Making Inferences Based on Your Sample

Presenters: Documenters:
KAY KAIZA F. INFANTE – MMEM-W SANITA R. AROMBO- MMPA-K
KRISTINE JOY A. ECHALUCE- MMEM- W RITZ ADRIAN BALLASOLA
IVY GRACE C. CASTILLO- MMEM-V LENY P. PAJUELAS – MMEM-V
ANNALIZA R. BUENVENIDA-MMEM-V MARIANITA LOBOS- MMEM-W
IN THIS CHAPTER, YOU WILL LEARN:
 About the use of inferential statistics to
determine whether the finding of a study is
unusual.
 The importance of the sampling distribution.
 How to carry out the hypothesis testing
process.
 When to reject or retain a null hypothesis and
the types of errors associated with each of
these decisions.
 The distinction between statistical significance,
effect size, and practical significance.
 How to compute, interpret, and report a one-
sample t test comparing a sample to a known
population value.
INFERENTIAL STATISTICS

In research we often want to go

beyond simply describing our sample
using measures of central tendency and
variability. Instead, we want to make
inferences about the qualities or
characteristics of the population that
our sample represents.
 INFERENTIAL STATISTICS
The statistics we use in this process are called
inferential statistics. When we use inferential
statistics we draw conclusions about a population
based on the findings from a single study with a
sample from the population.
We do not have to repeatedly conduct the
same study with different samples in order to
learn about the same population.
Of course, replication is not a bad idea but we
want to build information about populations, and
inferential statistics allows us to accomplish this
goal more quickly than conducting the identical
study multiple times with samples from the same
population.
Inferential VS Descriptive Statistics
INFERENTIAL DESCRIPTIVE
Population Sample
Parameter Statistics
Mu (µ) Mean (M)
Sigma (σ) Standard Deviation (SD)
 In inferential statistics we are interested in variables
or factors in populations rather than solely in the
information we have gathered from the sample we have
in our study.

The descriptive statistics you are familiar with

(means, standard deviations, etc.) are called parameters
when they are calculated from a population.

The mean of a variable for a population is called mu

(μ), and the standard deviation is called sigma (σ). We
do not often have these exact values (mu or sigma) as
we rarely have scores for an entire population.

Generally, our population is theoretical and

considered to be represented by our sample.
 Thus, inferential statistics use descriptive statistics
from our sample to make assumptions about a particular
population from which our sample is drawn.

When we use inferential statistics, we ask the

question: Is our sample representative of our population?
In other words, does our sample seem to belong to the
population or is it different enough that it likely
represents a totally different population?
Inferential statistics
 Statistical analysis of data gathered from
a sample to draw conclusions about a
population from which the sample is
drawn.

Parameters
 Statistics from a population.

Mu (μ): Population mean.

Sigma (σ): Population standard deviation.

Probability Theory
Inferential statistics are based on probability
theory, which examines random events such as
the roll of dice or toss of a coin. When random
events are repeated, they then establish a
pattern.

For example, we could toss a coin multiple

times to see the number of heads and tails that
result. We would expect to get heads
approximately 50% of the time and tails
approximately 50% of the time, especially if we
toss the coin a large number of times.
 In probability theory each individual random event
(in our example the outcome of one coin toss) is assumed
to be independent, which means that the outcome of each
coin toss is not affected by the outcome of previous tosses.

Another way to say this is that each time we toss a

coin, the outcome of our toss is not affected by whether
we got a head or a tail on
the last or earlier coin tosses.

The pattern that results from repeating a set of

random events a large number of times is used in statistics
to determine whether a particular set or sample of random
events (e.g., 8 heads in 10 coin tosses) is typical or
unusual.
 In research, we pay attention to the likelihood of
scores or findings occurring and we use probability theory
to help us determine whether a finding is unusual. An easy
way to understand how probability theory aids our
decision making is to think about the coin toss. If we flip a
coin 100 times and we get 45 heads and 55 tails, we
usually do not take note of the event.

If, however, we flip the coin 100 times and get 10

heads and 90 tails, we begin to wonder if our coin is
weighted or different in some way. We expect to get close
to the same number of heads and tails (in this case, 50
heads and 50 tails out of 100 tosses) and assume that
results that differ a great deal from this expectation may
indicate something unusual about the coin. But how
different from our expectation do our results need to be
for us to believe that there is something unusual about the
coin? We may begin to be suspicious if we get 20 heads
and 80 tails but that result could occur even with a
“regular” coin.
 Probability theory allows us to determine
how likely it is that our coin toss results would
have occurred in our particular circumstances. In
other words, what is the probability that we
would get 45 heads and 55 tails if we tossed a
coin 100 times?
The probability of our results is based on a
theoretical distribution where a task (flipping the
coin 100 times) is repeated an infinite number of
times and the distribution shows how frequently
any combination (of heads and tails) occurs in a
certain number of repetitions (100 flips of a coin).
We find out that we are likely to get 45 or
fewer heads 18.4% of the time when a coin is
tossed 100 times. (Of course, the same coin
would need to be tossed in the same way each
time in order for our results to be reliable.)
 If, however, we flip the coin 100 times and
get 10 heads and 90 tails, we begin to wonder if
our coin is weighted or different in some way.
We expect to get close to the same number of
heads and tails (in this case, 50 heads and 50
tails out of 100 tosses) and assume that results
that differ a great deal from this expectation
may indicate something unusual about the coin.
But how different from our expectation do our
results need to be for us to believe that there is
something unusual about the coin? We may
begin to be suspicious if we get 20 heads and
80 tails but that result could occur even with a
“regular” coin.
 In research, we pay attention to the
likelihood of scores or findings occurring and
we use probability theory to help us determine
whether a finding is unusual. An easy way to
understand how probability theory aids our
decision making is to think about the coin toss.
If we flip a coin 100 times and we get 45 heads
and 55 tails, we usually do not take note of the
event. If, however, we flip the coin 100 times
and get 10 heads and 90 tails, we begin to
wonder if our coin is weighted or different in
some way.
 We expect to get close to the same number
of heads and tails (in this case, 50 heads and 50
tails out of 100 tosses) and assume that results
that differ a great deal from this expectation may
indicate something unusual about the coin. But
how different from our expectation do our results
need to be for us to believe that there is
something unusual about the coin? We may
begin to be suspicious if we get 20 heads and 80
tails but that result could occur even with a
“regular” coin.
 Probability theory allows us to determine
how likely it is that our coin toss results would
have occurred in our particular circumstances.
In other words, what is the probability that we
would get 45 heads and 55 tails if we tossed a
coin 100 times?
The probability of our results is based on a
theoretical distribution where a task (flipping
the coin 100 times) is repeated an infinite
number of times and the distribution shows
how frequently any combination (of heads and
tails) occurs in a certain number of repetitions
(100 flips of a coin).
We find out that we are likely to get 45 or
fewer heads 18.4% of the time when a coin is
tossed 100 times. (Of course, the same coin
would need to be tossed in the same way each
time in order for our results to be reliable.)
Sampling Distribution VS
Frequency Distribution
In order to decide whether an outcome is
unusual, we compare it to a theoretical sampling
distribution.
In Chapter 5, we used the term frequency of
scores to describe distributions composed of
individual scores. In contrast, a sampling
distribution is composed of a large number of a
specific statistic, such as a mean (M), gathered
from samples similar to the one in our study.
 Each “score” on the distribution represents
the findings from tossing a coin 100 times. The
coin is tossed many hundreds of times and a
distribution is then built from all these results.

Thus, we have a large number of scores from

samples of 100 tosses, which give the
distribution the name sampling distribution.

Of course, no one actually tosses a coin

thousands of times—statisticians build the
distribution based on probability.
 To sum up this section:

Inferential statistics allow us to

determine whether the outcome
of our study is typical or unusual
in comparison to our sampling
distribution.
HYPOTHESIS TESTING

• The process of determining

the probability of obtaining a
particular result or set of
results.
Two Types Of Hypothesis
1. NULL HYPOTHESIS
 A prediction of no difference
between groups; the hypothesis
the researcher expects to reject.

2. ALTERNATIVE HYPOTHESIS
 A prediction of what the
researcher expects to find in a
study.
REJECTING A NULL
HYPOTHESIS?
Region Of Acceptance:
 Area of sampling distribution
generally defined by the mean
±2 SD or 95% of the distribution;
results falling in this region imply
that our sample belongs to the
sampling distribution defined by
the Ho and result in the
researcher retaining the Ho.
Region Of Rejection
 The extreme 5% (generally) of a
sampling distribution; results
falling in this area imply that our
sample does not belong to the
sampling distribution defined by
the Ho and result in the
researcher rejecting the Ho and
accepting the Ha.
ONE-TAILED TEST
VS.
TWO-TAILED TEST
 One Tailed Test
 A hypothesis stating the
direction (higher or lower) in
which a sample statistic will
differ from the population or
another group.

 Also known as the directional

 Two Tailed Test
 A hypothesis stating that results
from a sample will differ from
the population or another group
but without stating how the
results will differ.

 Also known as the non directional

 Critical value
 The value of a statistic that
defines the extreme 5% of a
distribution for a one-tailed
hypothesis or the extreme 2.5%
of the distribution for a two-
tailed test.
Setting the Criterion Level (p)

 Criterion level
The percentage of a sampling
distribution that the researcher
selects for the region of rejection;
typically researchers use less than 5%
(p < .05).
 ERRORS OF
HYPOTHESIS TESTING

 Type I Errors

 Type II Errors
 TYPE I ERROR
 The probability of rejecting a true
H0.
 defined by the probability of the
significance level of your findings.
 TYPE II ERROR
 The probability of incorrectly
retaining a false Ho.
 REDUCING THE CHANCE
OF A TYPE I ERROR

 make your study less vulnerable to

a Type I error by changing your
criterion level.
 Most of the time social scientists
use the p < .05
 REDUCING THE
CHANCE OF A TYPE II
ERROR
1. Sample size
(larger sample size = more power)

 In order to have power, you must have

enough members of the population
represented in your sample for you to feel
confident that the results you found in your
study are in fact the ones that exist in the
population.

 The larger the sample size the more confident

we can be that the mean of our sample
approaches the mean of the population from
which the sample is drawn.
2. Amount of error in the research design
(less error = more power)

 Within-groups variance (or error variance):

 The differences in your sample measure that
are not accounted for in the study.

 Homogeneity of the sample:

 The degree to which the members of a sample
have similar characteristics.

 Sensitivity
 The ability of a measurement instrument to
detect differences.
3. Strength of the effect
(stronger effect = more power)

 The strength of the effect refers to the

magnitude or intensity of a pattern or
relationship, and increases the likelihood that
the pattern or relationship will be detected.

 The strength of the effect is something that

we measure in research, and as such you do
not have as much ability to directly impact it
as you do with sample size and error.
 Measures we can use to
interpret the meaning and
importance of our findings

1. Statistical significance

2. Effect Size

3. Practical Significance
1. Statistical significance

 Is the claim that a certain

conclusion that’s drawn from a
data set probably didn’t occur
randomly and is instead likely
to have originated because of
a specific cause.
 To test for statistical significance,
perform these steps:

1. Decide on an alpha level. An alpha level is

the error rate you are willing to work with
(usually 5% or less).

2. Conduct your research. For example, conduct

a poll or collect data from an experiment.

3. Calculate your statistic. A statistic is just a

piece of information about your sample, like a
mean, mode or median.

4. Compare the statistic you calculated in Step 3

with a statistic from a statistical table.
2. Effect Size

 Tells the magnitude or strength of the

effect of a variable.
 One of the easiest types of effect size
is to understand the percentage of
variability in one variable (the dependent
variable), which is accounted for by the
independent variable (in the case of
experiments) or which is accounted for by
the relationship with another variable (in
the case of correlations).
Sample Interpretation of Effect Size:

Effect Size Range Interpretation

1–4% Weak
9 – 25% Moderate
25 – 64% Strong
3. Practical Significance

 Refers to the usefulness of our results

or findings from our study.
 Practical significance is usually
important to study after statistical
significance, as it's the only way to
exclude the possibility of a large
enough sample affecting the results
and identifying otherwise
meaningless effects.
 There are several ways in which
statistically significant data can be
applied in practical situations:

1. By comparing a sample to a meaningful

reference group.

2. By setting confidence limits.

3.By having the practical experience to detect

faulty data.
 COMPARING YOUR
SAMPLE TO A KNOWN
OR EXPECTED SCORE
One-sample t test

 An inferential statistic that

compares a sample mean to a
known population mean
The formula of one – sample t- test can be
derived by using the following steps:

1.Determine the observed sample mean, and the

theoretical population means specified. Where M
equals the mean of our sample; µ equals the
mean of the population.

2.Determine the standard deviation of the

sample, and it denotes by s.
Another way to consider the null
hypothesis is as an equation:

H0 : µ – M = 0

In other words, if there is no

difference between the sample and
population means, then subtracting
the mean from mu (µ) will equal
zero
Alternative Hypothesis (Ha)

We can express our alternative

hypothesis as either:
• a directional (one-tailed)
hypothesis; or
• a nondirectional (two-tailed)
hypothesis.
 Directional (One-tailed) or a
nondirectional (two-tailed) hypothesis
in numerical terms:

Ha : M > µ

where M equals the mean of our

sample; µ equals the mean of the
population.
Calculating a One-Sample t Test
Assumptions of the one-sample t
test include:

 Interval or ratio data

 Sample size of 30 or less

 Normally distributed population

 Availability of the population mean (µ)

 Calculating a one-sample t test is a
very similar process to calculating a z
score (that you learned about in
Chapter 5).
Recall that a z score represents the
number of standard deviation units a
score is from the mean
 It is computed by subtracting a
score from the mean and dividing
that difference by the standard
deviation of the sample.
 With the one-sample t test, we find the
difference between the population mean
and our sample mean. We then divide
that difference by the standard deviation
of the sampling distribution of means
which is called the standard error of the
means (σx ). Because σx is usually not
available, we use the estimated standard
error of the means (SDx ) to calculate
the t.

The subscript x denotes that we are

estimating the standard deviation of a
distribution of means rather than raw
scores.
 The subscript x denotes that we
are estimating the standard
deviation of a distribution of
means rather than raw scores.

To calculate SDx :

where SD is the sample standard

deviation; N is the sample size.
The definitional formula for a one-
sample t test is:

where M is the mean of the sample; µ is

the population mean; SDx is the estimated
standard error of the means;
SD is the standard deviation of the
sample; and N is the sample size.
Standard error of the means (σx ):

 Standard deviation of the

sampling distribution of means.

Estimated standard error of the

means (SDx ):

 Estimated standard deviation of

the sampling distribution of
means that is used to calculate
the t test.
 We use degrees of freedom (df) as an estimate of
sample size.
For a t test, df is equal to N – 1.
In the critical t value chart, we use df to find the value
that our calculated t value must equal or exceed in order
to reject the H0.

Degrees of freedom is formally defined as

the number of scores that are free to vary in a sample.

For example, if you think of three

scores that add up to 10, and the first two scores are 5
and 2, what must the third score be?
If you said 3, you are correct. The first two scores could be
any value but in order to add to 10, the third score in our
example must be 3. Thus, for our small group of scores,
df = 3– 1 = 2.
In general, we lose one df in any group of scores.
THANK YOU!!!