Experiments: Method and Methodology Mícheál Ó Foghlú Executive Director Research TSSG, WIT [email protected]@tssg.org March 2009.

Experiments: Method and Methodology

Mícheál Ó Foghlú

Executive Director Research

TSSG, WIT [email protected]

March 2009

Copyright © Mícheál Ó Foghlú 2009

Revised Schedule

Mon 12th Jan Thomas Magedanz - Guest Lecture on IMS [DONE] Wed 14th Jan Presentations [DONE] Wed 21st Jan Presentations [DONE] Wed 28th Jan IPv6 Summit (Dublin Castle) [DONE] Wed 4th Feb EMPTY Wed 11th Feb Session 01 [DONE] Wed 25th Feb EMPTY Wed 4th Mar Session 02 [DONE] Wed 11th Mar Session 03 [DONE] Wed 18th Mar EMPTY Wed 25th Mar EMPTY Wed 1st Apr Session 04 [Today] Wed 22nd Apr Session 05

Sessions 01-05 to be delivered by Mícheál Ó Foghlú


Schedule Detail

01 What is research?

– Philosophy, Epistemology, Methodology and Method 02 How to write academically?

– Some simple language rules– Some simple structure rules

03 What’s the big deal with plagiarism?– Bibliographies, references and citations, …– Doing it in Word– Doing it with other tools like LaTeX/BibTeX

04 Results - how to do experiments– Support tools: simulation, data analysis, …

05 Discussion


Structure

Experimental Design (basics) Statistical Analysis (basics)

Experimental Design

How to conduct a valid experiment.

http://www.slideshare.net/mrmularella/experimental-design


A Good Experiment

Tests one variable at a time. If more than one thing is tested at a time, it won’t be clear which variable caused the end result.

Must be fair and unbiased. This means that the experimenter must not allow his or her opinions to influence the experiment.

Does not allow any outside factors to affect the outcome of the experiment.


A Good Experiment

Is valid. The experimental procedure must test your hypothesis to see if it is correct.

If the procedure does not test your hypothesis, the experiment is not valid and the data will make no sense!

Has repeated trials. Repeating the trials in the experiment will reduce the effect of experimental errors and give a more accurate conclusion.


Variables

A variable is anything in an experiment that can change or vary.

It is any factor that can have an effect on the outcome of the experiment.

There are three main types of variables.


3 Kinds of Variables

Independent Variable (IV)– something that is intentionally changed

by the scientist– What is tested– What is manipulated– Also called a “Manipulated Variable”– You can only change ONE variable in an

experiment!!!



Independent Variable (IV)

To determine the independent variable, ask yourself:

“What is being changed?”

Finish this sentence…

“I will change the _____________”


Independent Variable

Levels of the IV These are different ways you will change the

independent variable

Example: Assume you are testing five brands of popcorn to see which has the most unpopped kernels.

The IV would be the different brands of popcorn. The five different brands would be the different

levels of the IV.



Dependent Variable (DV)

– something that might be affected by the change in the independent variable– What is observed and measured– The data collected during the investigation– Also called a “Responding Variable”



Dependent Variable (DV) To determine the dependent variable, ask yourself:

“What will I measure and observe?”


“I will measure and observe ________________”


Dependent Variable

Operational Definition: Define exactly how the dependent variable

will be measured.

Example: Assume your DV in an experiment is “plant growth.” How will you measure this?! It could be…

Height (cm), mass (g), # of leaves, etc. Be specific and include all necessary units!



Controlled Variable (CV)– a variable that is not changed and kept the same– Also called constants– Allows for a “fair test”– NOT the same as a “control”!!– Any given experiment will have many

controlled variables



Controlled Variable (CV)

To determine the controlled variables, ask yourself:

“What should not be allowed to change?”


“I will not allow the ______________ to change.”


Control

A group or individual in the experiment that is not tested, but is used for comparison as a reference for what “normal” would be like.

Not all experiments have a control (though all experiments have controlled variables).

Example: If you tested different pollutants to see their affect on plant growth, the control would only receive water.


Example

Students of different ages were given the same jigsaw puzzle to put together.

They were timed to see how long it took to finish the puzzle.


Identify the variables in this investigation!


What was the independent variable?

Ages of the students

– Different ages were tested by the scientist


What was the dependent variable?

The time it to put the puzzle together

– The time was observed and measured by the scientist


What was a controlled variable?

Same puzzle

– All of the participants were tested with the same puzzle.

– It would not have been a fair test if some had an easy 30 piece puzzle and some had a harder 500 piece puzzle.


Another Example:

An investigation was done with an electromagnetic system made from a battery and wire wrapped around a nail.

Different sizes of nails were used.

The number of paper clips the electromagnet could pick up was measured.


What are the variables in this investigation?


Independent variable:

Sizes of nails

– These were changed by the scientist.

– They used different sizes of nails in their experiment to see what effect that would have.


Dependent variable:

Number of paper clips picked up

– The number of paper clips were observed and counted (measured)


Controlled variables:

Battery, wire, type of nail

– None of these items were changed

– They had used the same battery, same wire, and same type of nail.

– Changing any of these things would have made it an unfair test.

Here’s another:


The temperature of water was measured at different depths of a pond.


Independent variable – depth of the water

Dependent variable – temperature

Controlled variables – same pond; same thermometer

Last one:


Students modified paper airplanes by cutting pieces off, adding tape, or adding paper clips to increase the distance thrown.


Independent variable – weight of plane, center of gravity, air resistance (depended on student choice-but only one was tested)

Dependent variable – distance thrown

Controlled variables – same plane design; same paper; same throwing technique


Now let’s take what we know about these variables and use them in an experiment!


We are going to test how many drops of water will fit on different sized coins.Let’s think about how we could test this.

– Identify the variables– What exactly will be changed? How

will it be changed?– What exactly will be measured?

How will it be measured?


Independent variable – size of the coin (penny, nickel, dime, quarter)

Dependent variable – amount of water held on coin (# of drops)

Controlled variables– Same eye dropper – Same water– Same side of coin (pick heads or tails)– Same technique (height/angle of dropper)

What are my variables?


Statistical Analysis

http://www.slideshare.net/sababutt/statistical-analysis-of-data-final-presentation


SIGNIFICANCE OF STATISTICS FOR ANALYSIS

AND RESEARCH


STATISTICS IS NECESSARY FOR ALL FIELDS OF LIFE REQUIRING RESEARCH AND DATA ANALYSIS

In all fields of life we have to analyze facts and interpret from these to make conclusions. The analysis needs statistics – to compare the qualities and quantities to help reach some conclusion, which will lead to decision making in business, government, industry etc and development of theories in science.


designing experiments and other data collection, summarizing information to aid understanding, drawing conclusions from data, and estimating the present or predicting the future.

In making predictions, Statistics uses the companion subject of Probability, which models chance mathematically and enables calculations of chance in complicated cases.

BIOSTATISTICS IS A DISCIPLINE THAT IS CONCERNED WITH:


SOME IMPORTANT DEFINITIONS


POPULATION AND SAMPLE

POPULATION: A population consists of an entire set of objects, observations, or scores that have something in common. For example, a population might be defined as all males between the ages of 15 and 18.

SAMPLE: A sample is a subset of a Population Since it is usually impractical to test every member of a population, a sample from the population is typically the best approach available.


PARAMETER AND STATISTIC

PARAMETER: A parameter is a numerical quantity measuring some aspect of a population of scores. For example, the mean is a measure of central tendency in a population.

STATISTIC: A "statistic" is defined as a numerical quantity (such as the mean calculated in a sample).


MEASURES OF CENTRAL TENDENCY

Mean (Arithmetic Mean)

Average value of a sample or population

Median

Middle value of sample or population

ModeThe value repeated most


The Arithmetic Mean or Mean is what is commonly called the average: When the word "mean" is used without a modifier, it can be assumed that it refers to the arithmetic mean. The mean is the sum of all the scores divided by the number of scores.

Formula of calculating Population Mean is:

μ = ΣX/N,

where μ = population mean, and

N = number of scores.

If the scores are from a sample, then the symbol X refers to the mean and n refers to the sample size, formula written as:

X = ΣX/n


Median: The median is the middle of a distribution: half the scores are above the median and half are below the median. The median is less sensitive to extreme scores than the mean and this makes it a better measure than the mean for highly skewed distributions.

5 3 4 2.5 6Mode: The mode is the most frequently occurring score in a distribution and is used as a measure of central tendency. The advantage of the mode as a measure of central tendency is that its meaning is obvious.

5 3 4 5 6


MEASURES OF DISPERSION

After measuring the central value i.e., mean, next is to know that to which extent this central value represents all values, that is, to know the scattering or dispersion of the data. There are certain measures which gives values of dispersion. The most important and widely used of these in research are: Variance Standard Deviation Standard Error of Mean


HYPOTHESIS TESTING

T testF testANOVACorrelationRegression


EXAMPLE OF DATA ANALYSIS

Comparison of Weight to Height Ratio expressed by Body Mass Index of a population. BMI is calculated as weight in Kg / Height in Meter2.

General surveys in USA and Europe showed that young population is overweight which is enhancing chances of diseases. We surveyed young female population of Punjab University for BMI. We measured BMI of 400 students randomly.


36.6620.2130.2929.3331.9727.5825.3326.9027.7427.0126.8222.6531.9030.8120.8425.1922.9828.6822.7322.8627.73

M-1M-2M-3M-4M-5M-6M-7M-8M-9M-10M-11M-12M-13M-14M-15M-16M-17M-18M-19M-20M-21

F-1F-2F-3F-4F-5F-6F-7F-8F-9F-10F-11F-12F-13F-14F-15F-16F-17F-18F-19F-20F-21

30.1128.0016.8738.9435.6332.6923.9225.5530.8743.4335.3419.6536.4534.3534.1538.8626.2829.5224.9929.7534.58

Subject No.

BMI Subject No.

BMI


We have two tables of data: one giving BMI of girls, other BMI of boys. These are long data tables.

Now, we have to analyze it to conclude something from this data . What we need, now?

We need a measure of central tendency to indicate average BMI to compare with other populations, between boys and girls and with the normal range.

The most common and useful measure for the purpose is the Arithmetic Mean. Arithmetic Mean is calculated by taking sum of all values and dividing it by No. of observations.

ARITHMETIC MEAN


SAMPLING ERROR

Then next, we have an average value but is this average representative of all values really. Is it possible that some values be very large and some very small? If it is so, the Mean is not representative of whole data. This is called sampling error because some students may have strong genetic tendency to being overweight, these values are somewhat different from population. This will make our result erroneous, i.e., our Mean does not represent all data.


EXAMPLE

We have four values - 2, 3, 4, 10 Mean = Sum of values / No of

Observations2 + 3 + 4 + 10 / 4

= 4.75This is far from three values in the data. This is because of a large value that exists in the data i.e. 10.


STANDARD DEVIATION

Now, we need some statistical measure that tell us how to rule out sampling error.

This is the standard deviation – measure to find how the individual values vary from the average value, i.e., Mean.


Standard Deviation of that Data

SD = s = ∑ (x – x) 2

n - 1

Descriptive Statistics from MINITAB

Variable N Mean Median StDev SE Mean

C1 4 4.75 3.50 3.59 1.80


T Test

Two Sample T-Test and Confidence Interval

Two sample T for BMI-F vs BMI-M N Mean StDev SE

MeanBMI-F 30 31.35 6.26

1.1BMI-M 21 26.96 4.11

0.90

95% CI for mu BMI-F - mu BMI-M: ( 1.5, 7.31)T-Test mu BMI-F = mu BMI-M (vs not =): T= 3.02

P=0.0040 DF= 48


Other Issues Covered

– Basics of experimental design– Basics of statistical analysis

Not covered - experimental design– Block structured design (e.g. Latin Squares)– Understanding experimental errors

Not covered - statistical analysis– Understanding the T Test and the large battery of

other tests (e.g. ANOVA)– Assumptions of tests (e.g. that observations are

normally distributed) and when it is invalid to use a test

– Discussion of significance So this talk just scratched the surface!

Experiments: Method and Methodology Mícheál Ó Foghlú Executive Director Research TSSG, WIT [email protected]@tssg.org March 2009.

Documents

mchel foghl slide

copyright mchel foghl

responding variable

variable cv

manipulated variable

independent variable

discussion slide

independent variable