What are the parametric tests ?

• Parametric tests are used only where a normal distribution is assumed.

• The most widely used tests are the t-test (paired or unpaired), ANOVA
(one-way non-repeated, repeated; two-way, three-way), linear
regression and Pearson correlation.

• Non-parametric tests are used when continuous data are not normally
distributed or when dealing with discrete variables.
Parametric Tests • The Most widely used are chi-squared, Fisher's exact tests, Wilcoxon's
matched pairs, Mann–Whitney U-tests, Kruskal–Wallis tests and
Prepared and compiled by Spearman rank correlation.
Cengiz Beşoğul, Instructor

Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 2

Parametric tests Parametric Tests Used in Statistical Data Analysis

Parametric tests used in Statistical Analysis
1. t-test for
Parametric Test Analysis Type Example
1. one-sample
Compare means between Is the mean systolic blood pressure (at baseline) for
2. independent two samples Two-sample t-test two distinct/independent patients assigned to placebo different from the mean for
3. dependent two samples (paired) groups patients assigned to the treatment group?
2. ANOVA for Compare two quantitative Was there a significant change in systolic blood
1. one-way non-repeated Paired t-test measurements taken from pressure between baseline and the six-month follow up
2. one-way repeated the same individual measurement in the treatment group?

3. two-way, three-way, Analysis of Compare means between

If our experiment had three groups (e.g., placebo, new
drug #1, new drug #2), we might want to know whether
3. Linear regression variance three or more
the mean systolic blood pressure at baseline differed
4. Pearson coefficient of correlation. (ANOVA) distinct/independent groups
among the three groups?
Estimate the degree of
Pearson coefficient Is systolic blood pressure associated with the patient’s
association between two
of correlation age?
quantitative variables
Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor
3 Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor
4

T-tests t-test for one sample

• The parameters given for one or two samples are compared by hypothesis • In t-test for one sample, test procedure tests whether the sample is drawn
test based on t-distribution. from the population specified.
• The procedure of t-test is simply summarized as following ; • The null hypothesis is stated on equality of sample mean and population
• t-test value for the sample data is computed and result compared with t- mean parameter. (or inequality (">", "<") status are tested)
critical(table) value. • The null hypothesis mostly claims the sample is belong to the population.
• The test is concluded in respect to the claim hypothized. Its alternative lies the null hypothesis.
• The test procedure is applied considering the hypothesis is focused on one • Some examples of t-test for one sample are following;
or two side of t-distribution shape. Ho: Mean of Sample = Mean of population
Ho : The sample observed is taken from the certain population
Ho : Mean of sample is ">" or "<" from a definite Mean of Population

Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 7 Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 8
 t-test for two samples
t-test for two samples • t-test for two samples are applied according whether they are independent
• In t-test for two samples, test procedure tests whether the mean of data or dependent.
groups differ for a definite value • If they are independent samples then the groups data must be observed
• The null hypothesis is stated on equality of means of samples. (or separately.
inequality (">", "<") status are tested) Assume the samples A and B are independent they are viewed by ven diagrams ;
• The null hypothesis mostly claims the averages of samples are not different A ∩ B = {} = Ø = Null A B
from each other. Its alternative lies the null hypothesis. The samples A and B are mutually exclusive samples
• Some examples of t-tests for two samples (Samples observed are A and B) • If they are dependent samples then the data must be observed(measured)
are following; as in pairs.
Ho: Mean of Sample_A = Mean of Sample_B Assume the samples A and B are dependent they are viewed by ven diagrams ;
A ∩ B ≠{} ≠ Null
Ho : The samples observed are taken from the certain population
The samples A and B are not mutually exclusive samples A B
Ho : Mean of sample_A is ">" or "<" from the Mean of sample_B
Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 9 Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 10

t-test for two independent samples t-test for two dependent samples
• If they are independent samples then the groups data comparison should • For dependent samples the data must be observed(measured) as in pairs.
be carried out according to whether variances of samples are equal or not. Assume the samples A and B are dependent they are viewed by ven diagrams ;
t-test value, standard deviation and degrees of freedom for the test are
computed with formulas ; A ∩ B ≠{} ≠ Null
The samples A and B are not mutually exclusive samples
• Variances are assumed equal ; A B
Variance _A = Variance_B

A B • t-test value and standard deviation of paired data are following ;

diff : difference assumed between means of samples and stnadard dev.of samples
Variance _A ≠ Variance_B
• Variances are assumed unequal ;

Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 11 Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 12

Analysis of variance(ANOVA) Analysis of variance(ANOVA)

• The Z test or t test are used to compare two populations, in comparing an
• The analysis of variance is method used for a comparison of experimental group with a control group.
different populations in studies which have several treatments or • But in ANOVA, a different probability distribution is used. It is F-distribution.
conditions.
• The anaysis procedure exams the variances derived within groups’ measures
• For example, we may want to compare mean scores from three or more and between groups’ parameters. The totals MSR and MSE are computed.
populations that represent three or more study conditions.
• So, the F-statistic is computed as ;
.... .... n populations SSR = the regression sum of squares
P1 P2 Pn-1 Pn SSE = the error sum of squares
SST = the total sum of squares (SST = SSR + SSE)
dfr = the model degrees of freedom (equal to dfr = k - 1)
The analysis of variance will enable us to extend the comparison to more than two groups. The hypothesis is dfe = the error degrees of freedom (equal to dfe = n - k - 1)
stated as ; k = the total number of groups (levels of the independent variable)
H0: µ1 = µ2 = µ3 = ... = µk ("all k population means are equal") n = the total number of valid observations
H1: At least one µi different ("at least one of the k population means is not equal to the others") dfT = the total degrees of freedom (equal to dfT = dfr + dfe = n - 1)
where ; µi is the population mean of the ith group (i = 1, 2, ..., k) MSR = SSR/dfr = the regression mean square
13
MSE = SSE/dfe = the mean square error
Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 14
Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor
 Pearson coefficient of correlation –Hypothesis-
Pearson coefficient of correlation
• The hypothesis of the significance test for correlation can be expressed in the
• The aim of such a study is to investigate any linear relationship following ways, depending on whether a one-tailed or two-tailed test is
requested:
between the two samples whether it exists. • Two-tailed significance test:
• The result is important not only for determining the existance of H0: ρ = 0 ("the population correlation coefficient is 0; there is no
a linear relationship exists but to measure the direction and association")
H1: ρ ≠ 0 ("the population correlation coefficient is not 0; a nonzero
strenght of such a relationship. correlation could exist")
• The sign of pearson coefficient the direction of the relationsship. • One-tailed significance test:
• The direction implies the upward(positive) or H0: ρ = 0 ("the population correlation coefficient is 0; there is no
association")
downward(negative) trend between the samples. H1: ρ > 0 ("the population correlation coefficient is greater than 0; a positive
• The pearson coefficient determines how the relationship is correlation could exist")
OR
strong. (evaluated between [-1;+1]) H1: ρ < 0 ("the population correlation coefficient is less than 0; a negative
correlation could exist")
where ρ is the population correlation coefficient.
Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 15 Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 16

The correlation coefficient for a sample

• The sample correlation coefficient between two variables x and y
is denoted r or rxy , and can be computed as:

End of chapter
where cov(x, y) is the sample covariance of x and y; var(x) is the sample
variance of x; and var(y) i
s the sample variance of y.

Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 17 Altınbaş University, Medicine School, Cengiz Beşoğul, Instructor 18