Done with lab work, which statistical analysis should i employ?

How to select a statistical tool for statistical analysis

1. Done with lab work, which statistical
analysis tool should I employ?
Abubakar Bilyaminu, PhD
Department of Pharmacology and Toxicology
UDU

2. What this session will cover
Knowing what statistical analysis
tool to employ after the curation of
your data

3. What this session will not cover
How to practically use and interpret
a statistical test and its output

4. Purpose of Statistical Analysis
Summarize and describe data
Draw inferences about populations
from sample data
Test hypotheses and make predictions

5. Types of Data
Qualitative Data:Categories or
labels (e.g., gender, color)
Quantitative Data:Numerical
values (e.g., height, age)

6. Descriptive Statistics
Measures of Central Tendency
Measures of Dispersion

7. Dangers of using wrong ST
Wrong interpretation of the findings
Affects the conclusion of the study.

8. Choosing a Statistical Test
Choice of a statistical test depends on:
Level of measurement for the dependent
and independent variables
Number of groups or dependent measures
The population parameter of interest
(mean, variance, differences between
means and/or variances)

9. Choosing a Statistical Test
Aim and objectives of the study
Type and distribution of the data
Nature of the observations
(paired/unpaired)

12. Mann U Whitney test
A nonparametric procedure that determines if
ranked scores (i.e., ordinal data) in two
independent groups differ.
It is also used to analyze interval or ratio scale
variables that are not normally distributed.

13. Wilcoxon Matched-Pair Signed Ranks Test
A nonparametric procedure that compares differences
between data pairs of data from two dependent samples

14. Kruskal Wallis H test
Nonparametric alternative of one way ANOVA test
The dependent variable is continuous (scale) but
not normally distributed or ordinal while the
dependent variable is categorical (nominal)

15. Assumptions for Kruskal Wallis test
The dependent variable should be measured
at the ordinal or continuous level e.g likert
scale or continuous variable (height, weight)
The independent variable should contain two
or more categorical independent groups e.g
male and female
Independence of observations (no
relationship btw the observations

16. Correlation
Extent to which two or more variable fluctuate
together
Positive correlation/negative correlation
Correlation can be perfect to no correlation
The values are from -1 to +1
Correlation does not imply causation
Pearson correlation and spearman ranked
correlation

17. When to use correlation
CRITERIA Pearson correlationSpearman Ranked order
One variable Scale Ordinal or scale
Another variableScale Ordinal or scale
Normality assumedYes No
Check for outlierYes No
Linear relationshipYes No
Denoted with r rho

18. Pearson correlation
Determine the strength and direction of the
linear relationship between two continuous
variables.
Test the hypothesis that there is no linear
relationship between two variables in the
population (i.e., zero correlation in the
population; r= 0).

19. Spearman ranked correlation
Measures the direction
Assumptions
The two variable are ordinal or scale
One variable must be monotonically related to the
other

20. Regression analysis
To assess strength of relationship
Btw a dependent and an independent variable
Predict the dependent variable from the
independent variable(s)
Types
Bivariate (two variables)
Multivariate (three or more variables)

21. Difference btw regression and correlation
CharacteristicsCorrelation Regression
Purpose Association /connection
btw 2 variables
Understandingthe link between
variables (one affects the other
Labels attachedNo clear labels (x and y can
be interchanged)
Clear distinction btw dependent
and independent variables
Inferential testCorrelation coefficientRegression coefficient,
intercept, t-statistics
Data Represented in a single
point
Represented in a line

22. THANK YOU