Simplified Essential

Concepts

Illustrated step-by-step
data analysis

The best free software

for your analysis
LEARN STATISTICS EASILY
 MASTER DATA ANALYSIS QUICKLY,
EFFORTLESSLY, AND WITH UNDENIABLE RESULTS

Discover our SIMPLE teaching method that will empower you to

analyze your data on your own in no time.

We cover ALL the essential steps and only what's truly necessary
for analyzing your data.

Built on the proven principle that it's entirely possible to

accurately analyze data without complex concepts or formulas.

This book will serve you well, even if you have no prior knowledge
of statistics. All 3 volumes are included in this offer.

GET YOURS BY
CLICKING HERE!

LEARN STATISTICS EASILY

 MORE INFORMATION ABOUT VOLUME III

The Volume III initiative is one-of-a-kind.

It allows us to incorporate topics suggested by

our followers in each new edition.

And those who have already purchased the

package will have lifetime access to these
updates.

Volume III covers more advanced topics than

those covered in the first two volumes.

As we plan to regularly update Volume III with

new topics, we encourage you to send us your
suggestions via our Instagram profile.

@learnstatisticseasily

LEARN STATISTICS EASILY

LEARN STATISTICS EASILY
 Applied Statistics: Data Analysis (Volume III - Advanced) / Alves,
Ana. - Learn Statistics Easily, 2023. 133 p. 2nd ed.

1. Statistics. 2. Biostatistics. 3. Learn Statistics Easily. 4. Applied

statistics. 5. Data analysis. 6. Inferential statistics. 7. Descriptive
statistics. 8. Graphs. 9. Sample size.

© 2023 Learn Statistics Easily

All rights reserved. This publication or any part of it may not be

reproduced or used without the copyright holder's express
permission.

Learn Statistics Easily

statisticseasily.com

LEARN STATISTICS EASILY

 PREFACE

Welcome to your ultimate guide to unlocking the

power of data analysis – quickly, easily, and
confidently.

This book presents a groundbreaking teaching

method that empowers you to independently
analyze your data with speed and precision.

We've distilled the essentials, providing only the

necessary information to conquer data analysis
without getting lost in complexities.

Say goodbye to intimidating concepts, formulas,

and tables. This guide is designed to benefit you,
even if your knowledge of statistics is limited.

Our innovative approach to "learning data analysis

quickly, easily, independently, and with
confidence" sets this book apart from the rest.

Let this guide be your invaluable companion as

you embark on the exciting data analysis journey.

LEARN STATISTICS EASILY

 MASTERING OUR
METHODOLOGY

(a) We distill only the most vital

concepts, making them
effortlessly understandable.

(b) Crystal-clear examples and

diagrams bring each concept to
life.

(c) Our algorithm for selecting

statistical analyses and graphs
is streamlined and
straightforward.

(d) We tackle the most

prevalent statistical analyses,
covering 99% of real-world
scenarios.

(e) Our detailed, step-by-step

instructions, paired with vivid
illustrations, make data
analysis a breeze to grasp.

(f) Experience the ultimate in

user-friendly, comprehensive,
and intuitive free statistical
software.

LEARN STATISTICS EASILY

 SUMMARY
GETTING STARTED:
ESSENTIAL KNOWLEDGE
Grasp the key concepts
in a simplified and
accessible manner.

TOP FREE STATISTICAL

SOFTWARE
Discover premier tools for data
analysis, graphing, spreadsheets,
and sample size calculations.

DESCRIPTIVE STATISTICS:
SUMMARY MEASURES
Dive into the most vital
measures for summarizing
and showcasing your data.

INFERENTIAL STATISTICS:
UNLEASHING DATA ANALYSIS
Learn to choose the
right analysis and
apply it with precision.

PICKING THE PERFECT

GRAPH (VOL. II)
Follow a step-by-step guide
to selecting and creating the
ideal graph for your data.

BONUS CONTENT & ADVANCED

TOPICS (VOL. III)
Delve into extra tips and
explore slightly more
sophisticated subjects.

LEARN STATISTICS EASILY

 TABLE OF CONTENTS - VOLUME III

CALCULATING SAMPLE SIZE

1. Why calculate sample size?..................................................11
2. Sample size calculation software........................................14
3. Key parameters of G*Power................................................15
4. Calculating the sample size.................................................20
5. Comparing unpaired groups...............................................24
6. Comparing paired groups....................................................41
7. Looking for relationship between variables......................69

HOW TO PERFORM A SAMPLE

1. Why should we use an appropriate sampling method....96
2. Basic concepts.......................................................................97
3. Types of sampling...............................................................100
4. Final considerantions on sampling...................................104
5. How to do in practice.........................................................107

EXTRA ADVANCED TOPICS

1. P-value: the return..............................................................112
2. Effect size.............................................................................118
3. Confidence interval.............................................................122
4. Test Power...........................................................................130

ADVANCED INFERENTIAL ANALYSIS

1. Generalized linear model.......................................upcoming

LEARN STATISTICS EASILY

 TABLE OF CONTENTS - VOLUME III

CHAPTER 1: CALCULATING SAMPLE SIZE

LOOKING FOR DIFFERENCES

BETWEEN UNPAIRED GROUPS
1. Independent samples t-test (Mann-Whitney U-test)..............25
2. One-way ANOVA (Kruskal-Wallis H-test)...................................30
3. Two-way ANOVA (factorial)........................................................35

LOOKING FOR DIFFERENCES

BETWEEN PAIRED GROUPS
4. Paired samples t-test (Wilcoxon signed-rank test)..................42
5. One-way repeated measures ANOVA (Friedman test)...........48
6. Two-way repeated measures ANOVA.......................................55
7. Mixed-design ANOVA..................................................................62

LOOKING FOR RELATIONSHIPS

BETWEEN VARIABLES
8. Pearson correlation (Kendall's Tau & Spearman's Rank).......70
9. Simple linear regression.............................................................74
10. Multiple linear regression........................................................78
11. Simple binary logistic regression............................................84
12. Multiple binary logistic regression..........................................91

LEARN STATISTICS EASILY

LEARN STATISTICS EASILY
learn statistics easily
 1. WHY CALCULATE SAMPLE SIZE?

When conducting a sample, we collect data from

only a portion of the elements that make up the
population.

This is because collecting data from all elements in

the population would be time-consuming and
expensive.

Moreover, collecting all this data would be

unnecessary if we follow the steps outlined here.

Why is defining an appropriate

sample size important?

An appropriate Unduly large Excessively

sample size samples small
results in a result in the samples
reduction of wastage of yield
the sampling time and unreliable
error. money. results.

LEARN STATISTICS EASILY 11

 Due to the sampling error — represented by the
difference between the true population value and
the sample result — a sample will never perfectly
represent the population.

= -
SAMPLE TRUE POPULATION SAMPLE
ERROR VALUE RESULT

This error hinders our ability to infer population

characteristics from the sample data.

Depending on its magnitude, this error can lead to

false conclusions.

Therefore, in a more superficial analysis, the

smaller the sampling error, the better.

How can we reduce the sampling error?

Firstly, we should always consider the best cost-

benefit balance. Then, using an adequate sample
size calculated beforehand, we should collect the
data using an appropriate sampling method
(Volume III: Chapter 2).

LEARN STATISTICS EASILY 12

 2. SAMPLE SIZE CALCULATION
SOFTWARE

G*Power: Statistical Power Analyses is a powerful

and free statistical tool for sample size calculation.

Aside from other functions, such as those related

to test power and effect size, G*Power is the ideal
software for computing the sample size of the
tests we cover in this book, including t-tests,
ANOVA, linear regression, and so on.

You can download G*Power from its official

website.

On the website, scroll down until you find the

Download section, choose the desired version,
either Windows or macOS, and download and
install G*Power.

LEARN STATISTICS EASILY 14

 Each sample size calculation in G*Power can be
divided into three simple steps:

1. Selection of the appropriate test

2. Inputting the parameters
3. Estimating the effect size
1

LEARN STATISTICS EASILY 21

 Group 1 Group 2 Group 3

CALCULATING THE SAMPLE SIZE

(ARE THESE UNPAIRED GROUPS DIFFERENT?)

5.1 INDEPENDENT SAMPLES T-TEST

5.2 ONE-WAY ANOVA

5.3 TWO-WAY ANOVA (FACTORIAL)

LEARN STATISTICS EASILY

 5.1 INDEPENDENT SAMPLES T-TEST

Example:
Test if there is a difference in height between male
and female individuals in an indigenous tribe.

Sample Size Calculation:

From a pilot study, we obtained 4 observations from each of the
2 groups and then estimated their mean and standard deviation.

1. When opening G*Power, click on Tests: Means: Two

independent groups.
1

LEARN STATISTICS EASILY 25

 2. Now we will fill in the Input Parameters according to the
following information:

Tail(s): Mark One if the test is one-tailed or Two if the test is two-
tailed. But when will the test be one-tailed or two-tailed?

If the alternative hypothesis (H1) is specific, for example, "the mean

of group 1 is greater than the mean of group 2," we use the one-
tailed test.

If the alternative hypothesis is general — without the initial

distinction of greater or smaller — for example, "the means are
different between the groups," we use the two-tailed test.

This hypothesis should be based on existing knowledge in the

study field. If you are unsure, keeping the Two-tailed option is
best.

α err prob (significance level): The probability of rejecting the

null hypothesis when it is true (Type I error).

Usually, the values used are 0.05 or 0.01. A significance level of

0.05, for example, indicates a 5% risk of concluding that there is a
difference when there is no real difference.

LEARN STATISTICS EASILY 26

 Power (1 – β err prob) (test power): The probability of rejecting
the null hypothesis if false is how much the test controls Type II
error.

Usually, an acceptable value is between 0.80 and 0.99.

The higher the test power, the better, but as we increase it, the
required sample size also increases.

Allocation ratio N2/N1: If you want the sample size of the two
groups to be the same, enter the value 1.

Enter the required value if you wish to have another allocation

ratio between the groups.

3. Effect size d: This measure represents the strength of the

difference detected between the tested groups.

Conventionally, d = 0.20 is considered weak, d = 0.50 is

considered medium, and d = 0.80 is considered strong.

To estimate d for our sample size calculation, click on Determine

=>, and a new tab will open. In this new tab, we need to enter
values for the Mean and SD σ (standard deviation), which we
obtained from the pilot study, for each of the two groups.

These summary measures can be easily calculated according to

the instructions in Volume I: Chapter 3: Topic 5.

After filling in, click on Calculate and transfer to the main

window. The d value appropriate for your data will be calculated,
and the Effect size d field will be filled in.

LEARN STATISTICS EASILY 27

 2

3
1

4. With all the Input Parameters filled in, we need to click

Calculate in the lower right corner to obtain the number of
samples we need to collect.

In this case, our calculated total sample size was 14 elements, 7 in

group 1 and 7 in group 2, which we must obtain randomly from
the population.

Thus, we will need this sample size to detect this effect size; and,
assuming it is correct, with a 5% probability of committing a Type I
error and 95% probability of not committing a Type II error.

LEARN STATISTICS EASILY 28

 NOTES

1. How can we obtain the two groups' mean and standard

deviation values to estimate the effect size?

Commonly, these values are estimated in two ways:

(1) through a pilot collection with a few samples or

(2) through other studies with similar populations, species, or

conditions.

Sometimes, we may have reason to directly assume the expected

effect size without estimating the mean and standard deviation
summary measures.

2. To calculate the sample size for the Mann-Whitney U test, a

non-parametric equivalent to the t-test:

Click on Tests: Means: Two independent groups: Wilcoxon

(non-parametric) in the first step.

Keep the Parent distribution option as Normal.

The rest of the steps are the same.

LEARN STATISTICS EASILY 29

 Group 1 Group 2 Group 3

CALCULATING THE SAMPLE SIZE

(ARE THESE PAIRED GROUPS DIFFERENT?)

6.1 PAIRED SAMPLES T-TEST

6.2 ONE-WAY REPEATED MEASURES ANOVA

6.3 TWO-WAY REPEATED MEASURES ANOVA

6.4 MIXED-DESIGN ANOVA

LEARN STATISTICS EASILY

 6.1 PAIRED SAMPLES T-TEST

Example:
Does medication use affect older women's blood pressure (BP)?
The BP of each participant will be measured twice, once
before and once after the intervention (medication).

Sample Size Calculation:

From a pilot study, we obtained BP measurements of 4 older
women before and after the intervention. We then calculated the
data's means, standard deviations, and correlation coefficient.

1. To start, open G*Power and click on Tests: Means: Two

dependent groups (matched pairs).
1

LEARN STATISTICS EASILY 42

 VARIABLE Y
VARIABLE X

CALCULATING THE SAMPLE SIZE

(ARE THESE VARIABLES RELATED?)

7.1 PEARSON CORRELATION

7.2 SIMPLE & 7.3 MULTIPLE LINEAR REGRESSION

7.4 SIMPLE & 7.5 MULTIPLE LOGISTIC REGRESSION

LEARN STATISTICS EASILY

 7.1 PEARSON CORRELATION

Example:
Testing whether there is a correlation between the
weight and height of members of an indigenous tribe.

Sample Size Calculation:

We obtained the weight and height of 6 individuals
from a pilot study and estimated the correlation
coefficient between these two variables.

1. Open G*Power and click on Tests: Correlation and

regression: Correlation: Bivariate normal model.
1
2

LEARN STATISTICS EASILY 70

learn statistics easily
 1. WHY SHOULD WE USE AN APPROPRIATE
SAMPLING METHOD?

We saw in the previous chapter that calculating the sample

size is essential for us to have more reliable results in our
analyses at the best possible cost-benefit.

However, suppose the sampling is inadequately done —

in that case, it is not enough to calculate the sample size
and obtain an adequate number of elements.

A poorly done sampling introduces biases in the sample,

increases the sampling error, and leads us to draw false
conclusions.

Sampling represents a process in which we select a part

of the elements that make up a population to characterize
it.

This process can be done in numerous ways.

Still, there are more appropriate ways to do it according

to the situation.

Thus, we will discuss sampling methods and how we

should conduct them correctly.

LEARN STATISTICS EASILY 96

 3. TYPES OF SAMPLING

There are two general types of sampling, non-

probabilistic and probabilistic.

In non-probabilistic sampling, there is a deliberate and

arbitrary selection of elements from the population.

It depends on subjective criteria of accessibility and

researcher judgment.

They are called, for example, convenience sampling, and

judgment sampling, among others.

Thus, for obvious reasons, we will not discuss these types

of sampling here.

In probabilistic sampling, the selection of elements is

entirely random, so each individual in the population has a
known probability of being part of the sample.

It includes the most rigorous and widely used selection

methods in scientific research.

"We must not forget that sampling must always be

PRECISE and REPRESENTATIVE of the population."

LEARN STATISTICS EASILY 100

 3.3 STRATIFIED RANDOM SAMPLING

Stratified sampling is a probabilistic sampling method

used when one or more criteria must be considered in the
population before selecting elements. This method
involves two steps:

1. Dividing the population into two or more groups (strata)

based on the characteristic of interest. 2. Randomly select
elements from each stratum proportionally to its size.

This method allows for better control of the

representativeness of different elements within the
population.

STRATIFIED
RANDOM SAMPLING
EXAMPLE: To select 6 people from a population of 18,
dividing the population into 2 strata based on sex.
Then, randomly choose 3 men and 3 women from
their respective strata, proportionally to their size.
POPULATION STRATA SAMPLE

LEARN STATISTICS EASILY 103

 DO YOU
PREFER DOGS
OR CATS?
THE
COOLEST DOG
CONTEST

In 1948, a Chicago newspaper erroneously

projected the next American president based on a
phone survey. They did not consider that, at that
time, only the upper class could have a telephone.
The golden lesson is to always ensure that your
sample represents the population.

LEARN STATISTICS EASILY 106

 5. HOW TO DO IT IN PRACTICE

We could do the drawings the old-fashioned way, like using a bag

or an urn in a bingo game. But obviously, we have more practical
and faster possibilities for this nowadays.

So, first, we need each population element to have a unique

identification and to be arranged in a spreadsheet. This
identification can be any number, name, code, social security
number, etc.

Remember that each element can be a person, an animal, a

medical record, a quadrant, a country, a city, a forest, a lagoon, a
Petri dish, a house, a purchase, etc.

For our example, let's suppose we have 200 names of patients

treated in a clinic, and we want to sample 20 for research.

1. Here, we have this example with the patients' names in

alphabetical order.

LEARN STATISTICS EASILY 107

learn statistics easily
 1. P-VALUE: THE RETURN

In Volume I, we presented a didactic definition of p-value,

but it does not accurately reflect its meaning.

Therefore, for initial contact with the topic, this

simplification contributes to its understanding.

Thus, as promised, we present a precise definition of the

p-value, which simultaneously requires greater attention
and abstraction.

When we perform an inferential hypothesis test, such as

chi-square, t-test, ANOVA, correlation, regression, etc., we
basically have two hypotheses:

NULL ALTERNATIVE
HYPOTHESIS (H0) HYPOTHESIS (H1)

The standard, simplest, An alternative, complementa-

that there is 'no difference ry state to H0, that there are
between the groups' or 'no 'differences between groups' or
relationship between 'relationships between
the variables'. the variables'.

Thus, the fundamental objective of any hypothesis test is

to define whether we will reject or not the null hypothesis
(H0) — and this definition will depend on two
fundamental factors:

LEARN STATISTICS EASILY 112

 Thus, in technical terms, the p-value can be precisely
defined as:

The p-value represents the probability of

obtaining a result equal to (or more
extreme than) the one obtained from our
data, assuming that the null hypothesis is
true.

If my test returned, for example, a p-value of 2%,

what does this mean?

If we consider H0 true, the probability of obtaining results

equal to (or more extreme than) ours would be only 2%.

However, since it is lower than α = 5%, we reject H0.

See next for a more detailed explanation.

p < 0.05
"We Are The
Champions"

LEARN STATISTICS EASILY 114

 It was a summer afternoon in Harpenden, England, in the
early 1920s.

A group of scientists — a statistician, a phycologist, and a

biochemist — had sat at a table for afternoon tea at the
Rothamsted Experimental Station.

The phycologist, Muriel Bristol, insisted that tea poured

over milk tasted different than when milk was poured over
tea.

Everyone questioned her — What could be the difference?

They couldn't conceive that there could indeed be any

difference in taste.

LEARN STATISTICS EASILY 115

 3. CONFIDENCE INTERVAL

Some concepts already discussed in Volume I are

essential for understanding the confidence interval, such
as population, sample, and sampling error.

POPULATION
samp
lin
POPULATION g
(parameter)

in
sample
fe (statistics)
hypothesis ren
testing ce
parameter
estimation

A population encompasses all elements.

On the other hand, a sample includes only a portion of

these elements.

Inferential statistics allow us to draw conclusions about

the population through the sample.

However, since a sample represents only a part of the

population, it will never be perfect.

Thus, the sampling error is born, represented by the

difference between the true population value and the
value obtained in the sampling.

LEARN STATISTICS EASILY 122

 Imagine we want to estimate the average weight of all
oranges in an orchard.

The best way to do this is by collecting a random sample

with an appropriate sample size and presenting the mean
and the confidence interval.

What factors can influence the size of the confidence

interval?

(1) The larger the sample size, the smaller the confidence
interval.

(2) The lower the variability within the population, the

smaller the confidence interval.

(3) The lower the confidence level (1 - α) — for example,

90, 95, or 99% — the smaller the confidence interval.

LEARN STATISTICS EASILY 124

 Now, back to the oranges.

In our sampling of 10 oranges, we obtained a mean

weight of 150 g with a confidence interval (95%) of 125 to
175 g.

But what does this mean?

To explain, we will perform 19 other random samples,

each with 10 other oranges randomly selected from the
orchard.

Thus, 95% of the 20 random samples we made will

contain the population mean of the weight of the
oranges in the orchard.

See the graph with an example of 20 samples, where only

one (5%) does not contain the population mean, as
expected.

LEARN STATISTICS EASILY 125

LEARN STATISTICS EASILY
 MASTER DATA ANALYSIS QUICKLY,
EFFORTLESSLY, AND WITH UNDENIABLE RESULTS

Discover our SIMPLE teaching method that will empower you to

analyze your data on your own in no time.

We cover ALL the essential steps and only what's truly necessary
for analyzing your data.

Built on the proven principle that it's entirely possible to

accurately analyze data without complex concepts or formulas.

This book will serve you well, even if you have no prior knowledge
of statistics. All 3 volumes are included in this offer.

GET YOURS BY
CLICKING HERE!

LEARN STATISTICS EASILY

Unlock the secrets to analyzing your data swiftly,
effortlessly, and confidently.

Our SIMPLE approach focuses on teaching you

precisely what you need to know to conquer data
analysis.

Leave behind the complexities of concepts,

formulas, and tables – this course proves that
accurate data analysis is achievable for everyone.

This accessible resource is tailored for those with

little or no prior knowledge of statistics.

Discover our unparalleled method for "fast, easy, and

confident data analysis" – a game-changer you won't
find anywhere else.

LEARN STATISTICS EASILY