Statistical inferences

Statistical inferences
• Using the data from the sample to provide information
about the population parameter
• The accuracy of the collected sample data would
enable to generalize the findings for the population of
interest.

• Common approaches:
1. Hypothesis testing
2. P-value
3. Confidence interval
Hypothesis testing
• Hypothesis testing is the process used to evaluate
the strength of evidence from the sample and
provides a framework for making determinations
related to the population
Steps for hypothesis testing
• Step 1: State the null and alternate hypotheses
• Step 2: Decide on the significant level, α
• Step 3: Compute/calculate the test statistic
• Step 4: Decision to reject (or fail to reject) the null
hypothesis
• Step 5: Interpretation
 Hypothesis testing
• Helps us to choose between 2 conclusions
1. The treatment works versus the treatment does not work
2. There is an association versus there is no association

• Null Hypothesis: a tentative position that there is no association/difference

between the 2 sets of values
• Mathematically, it would have an equal sign
H0: µ1 = µ2
H0: µ1 ≥ µ2
H0: µ1 ≤ µ2
• Alternative hypothesis: a tentative position that there is an association or
difference between the 2 sets of values
• Mathematically, it would not have an equal sign
HA: µ1 ≠ µ2
HA: µ1 < µ2
HA: µ1 > µ2

We generally try to disprove the null hypothesis

Step 1: State the Null and alternate
hypotheses
Can we conclude that a certain population mean is not
50? (this is 2-tailed)
H0: µ = 50
Ha: µ ≠ 50
Can we conclude that a certain population mean is
greater than 50? (this is 1-tailed to the right)
H0: µ ≤ 50
Ha: µ > 50
Step 2: Decide on the significance level
• The level of significance (α) is a probability of rejecting
a true null hypothesis.
• Denoted as α, it marks the critical portion of the
normal distribution curve, beyond which will be the
rejection/significant region
• We normally take this value as 0.05
Step 2: Decide on the significance level
• 1 tailed or 2 tailed?
Step 2: Decide on the significance level
Step 3: Compute the test statistic
Step 4: Decision
• If the computed value of the test statistic falls in the
rejection region, it is said to be significant. If not,
then we fail to reject the null hypothesis
Step 5: Interpretation
• Based on the above steps, we provide the
interpretation as: “we have sufficient evidence to
prove that…..”
 P-value
• The p value for a test may be defined as the smallest
value of α for which the null hypothesis can be
rejected.

• Typically, an estimate that has a p value < 0.05 is considered

to be “statistically significant” or unlikely to occur due to
chance alone.
 Confidence interval
• The probability that a population parameter will fall between a
set of values for a certain proportion of times.
• Analysts often use confidence intervals that contain either 95%
or 99% of expected observations.

Confidence interval = (Estimator) ± Z x (standard error)

Example
• In a study, a sample of 100 subjects were selected by
simple random sampling. Their mean age was found
to be 54.85. Perform a test of significance to
determine the likelihood that such a sample mean
comes from a population whose mean is 53, given
that σ = 5.50 and α = 0.05.
Example
• Step 1: State the H0 and Ha
H0: µ = 53
Ha: µ ≠ 53
• Step 2: Determine the α = 0.05
Example
• Step 3: Compute the test statistic

Z = 1.85/0.55
Z = 3.36
Example
• Step 4: Decision

Z = 3.36 lies in the rejection region, hence it is

significant and we reject the H0 (and accept Ha)
Example
• Step 5:Interpretation

We have sufficient evidence to conclude that the

sample mean DOES NOT come from a population
whose mean is 53
Determining the p-value…
• Zcalc = 3.36
• The probability value (area under the curve) comes out as = 0.4996
• Subtraction: 0.5 – 0.4996 = 0.0004
• The p-value being < 0.05, hence we reject H0.
Determining the confidence interval…

Sample mean = 54.85

Standard deviation = 5.50
n = 100
z = 1.96 (at 95%)

95% CI = (53.772, 55.928)

As this range does not capture the value stated at H0,
we reject H0
 T-test
Chi-square test
T-test
• Discovered by William Gossett in 1906
• A method of testing hypothesis about the mean of small sample
drawn from a normally distributed population when the standard
deviation for the population is unknown.
• The formula,
• The t-distribution has a single parameter called the number of
degrees of freedom (df)—this is equal to the sample size minus 1.
df = n-1

• For large samples (> 30) the sample variance pretty much
approximates the population variance. In this situation, for all
practical reasons, the t-statistic behaves identically to the z-statistic.
Example
• The level of phosphate, in mg/dl in the blood of a patient undergoing
dialysis treatment was measured on six consecutive visits.
5.6, 5.1, 4.6, 4.8, 5.7, 6.4.

Calculate the sample mean = ?

Calculate the sample standard deviation = ?
Example
Calculate the sample mean = 5.4 mg/dl
Calculate the sample standard deviation = 0.67 mg/dl

• Test the hypothesis that the sample belonged to the population

where its mean phosphate levels is 4.0 mg/dl. Assume α = 0.01, and
the t-test value at df = 5, is 4.032
• Step 1: State the H0 and Ha
H0: µ = 4 mg/dl
Ha: µ ≠ 4 mg/dl (2-tailed)

• Step 2: Determine the α = 0.01. T-table shows that the 2 tailed t-test
value at df = 5, is 4.032
• Step 3: compute test statistic

t = (5.4-4.0)/(0.67/√6) = 5.12

• Step 4: Decision
As the tcal>tobs, we reject H0

• Step 5: Interpretation
We have sufficient evidence to conclude that the sample
DID NOT belonged to the population whose mean
phosphate level is 4 mg/dl at 0.01 level of significance.
Types of T-test
• One sample t test – we have only 1 group; want to
test against a hypothetical mean.

• Independent samples t test – we have 2 means, 2

groups; no relation between groups. Eg, mean heights
between male and female students of a college

• Paired t test – It consists of samples of matched pairs

of similar units or one group of units tested twice. Eg,
mean Hb among females before and after
administering intervention X.
Comparison of 2 Sample Means
• Independent samples’ T test
• Assumes normally distributed
continuous data.

T value = difference between means

standard error of difference

• T value then looked up in Table to

determine significance
Paired T Tests
• Uses the change before and
after intervention in a single
individual
• Reduces the degree of
variability between the
groups

T value = difference between means

standard error of difference

• Given the same number of

patients, has greater power to
detect a difference between
groups
Chi-square tests
• Applicable for qualitative (categorical data)
• Performed to determine associations between 2 (categorical)
variables
• Chi-square test compares the observed frequencies
with the expected frequencies.
• Formula,

𝑹𝑻 ×𝑪𝑻
• Expected value =
𝑻𝑻
• df = (r-1)(c-1)
Example
• Results from the Isoniazid drug trial after 6 months of follow up
yielded the following:

Dead Alive TOTAL

Placebo 21 110 131
Isoniazid 11 121 132
TOTAL 32 231 263

• Test at α = 0.05, if an association is present between use of isoniazid

and mortality.
Chi-Squared (2) Test-an example
Step-1: Hypotheses
H0: There is no association between isoniazid and mortality
HA: There is an association between isoniazid and mortality

Step-2: State the α (α= 0.05). Chi-square value at α= 0.05

and df = 1 is 3.84

Step-3: Calculations
O E O-E (O-E)2 (O-E)2/E
21 15.9 5.1 26.0 1.6
11 16.1 -5.1 26.0 1.6
110 115.1 -5.1 26.0 0.2
121 115.9 5.1 26.0 0.2
3.7

Chi-square = 3.7
Chi-Squared (2) Test-an example
Step-4: Decision
As Chi-squarecal < Chi-squaretab we fail to reject H0

Step-5: Interpretation
We have no evidence to conclude that there is an
association between isoniazid and mortality at α = 0.05
Errors in statistical inferences
Errors in statistical inferences
• 2 types:
• Type 1 error (α error)
reject H0 when H0 is true
• Type 2 error (β error)
failed to reject H0 when H0 is false
Type 1 error
• It means rejecting the null hypothesis when it’s
actually true.
• It means concluding that results are statistically
significant when, in reality, they came about purely by
chance or because of unrelated factors.

Example:
You decide to get tested for COVID-19 based on mild
symptoms.

Type 1 error: the test result says you have coronavirus,

but you actually do not.
Type 2 error
• It means not rejecting the null hypothesis when it’s
actually false.
• It means failing to conclude there was an effect when
there actually was.
Example:
You decide to get tested for COVID-19 based on mild
symptoms.

Type 2 error: the test result says you don’t have

coronavirus, but you actually do.
How to avoid these errors?
• Type 1 error:
• The α level is usually set at 0.05 or 5%.
• Lowering the α could assist in reducing the probability of type 1 error.
How to avoid these errors?
• Type 2 error:
• The risk of making a type 2 error is inversely related to the
statistical power of a test.
• Statistical power is determined by:
• Effect size: Larger effects are more easily detected.
• Measurement error: Systematic and random errors in recorded
data reduce power.
• Sample size: Larger samples reduce sampling error and increase
power.
• Significance level: Increasing the significance level increases
power.
• Increasing the sample size or the significance level will
reduce the probability of type 2 error.