Sr Dr Mona Isa FSPU UiTM Shah Alam 6/3/2016

RESEARCH METHODOLOGY &

MODULE 4: DATA ANALYSIS
COMPUTER (RES510)
DR SR MONA ISA •Descriptive Analysis
DEPARTMENT OF ESTATE MANAGEMENT
FACULTY OF ARCHITECTURE, PLANNING AND SURVEYING
•Correlation Analysis
(FAPS), UiTM SHAH ALAM •Regression Analysis and Multiple Regression
Analysis
•Time Series Analysis and Applications
*Slides are sourced from various sources on the net and RM
books

DATA ANALYSIS STASTISTICAL ANALYSIS

 Data analysis is a process of editing, Parametric Test
coding, processing and analysing the
raw data for purpose of building a Non-Parametric Test
sort of empirical model where
relationships are carefully brought out
so that some meaningful inference
can be drawn

STATISTICAL ANALYSIS AND

MANIPULATION Levels of Measurement
 Measurement is the process of assigning numbers to
Parametric Non-Parametric
variables.
Data normal Data is not normal
One-sample t Wilcoxion
 For example: Males and females in a study.
Paired-sample t Friedman
Males would be assigned as 1 and females assigned as 2.
Independent T-test Mann-Whitney
ANOVA/MANOVA Kruskal-Wallis
Pearson Spearman rho  Every variable in research study that is assigned a specific
Single/Multiple regression number must be similar to every other variable assigned that
number.

1
Sr Dr Mona Isa FSPU UiTM Shah Alam 6/3/2016

Levels of Measurement Ordinal

 Nominal- aka categorical, naming or classifying. Either does  Used to show relative rankings of variables or events.
or does not have the characteristic.  Ranks in order from high to low, but does not indicate
how much higher or how much lower.
 Lowest level of measurement and allows for the least amount  Intervals are not necessarily equal and there is no
of statistical information. absolute zero.
 Limited in the amount of mathematical manipulation
 Examples- gender, marital status, religious affiliation. possible.
 Examples- class rank, levels of wellness, levels of height.
 Can you think of one?

Interval Ratio
 Shows rankings of events or variables on a scale with equal  Shows rankings of events or variables on scales with equal
intervals between. interval and absolute zero.
 Zero point remains arbitrary and not absolute.  Most often used in physical sciences.
 Allows for more mathematical manipulation of data.  Highest level of measurement, allows for most manipulation
 Examples- test scores and temperature on a Fahrenheit scale. of data.
 Number represents the actual amount of the property the
object possesses.
 Example- height, weight, pulse and BP.

NOIR Descriptive Statistics

 No
Procedures that allow researchers to
 Oil
 In
describe and summarize data you
 Rivers definitely know (describes the sample).

Examples: Demographics data.

2
Sr Dr Mona Isa FSPU UiTM Shah Alam 6/3/2016

Descriptive Statistics Normal Distribution

Measures of central tendency are used to describe the  Most important curve (Bell-shaped).
pattern of responses among a sample.
 Mean- most frequently used average, add up
numbers (sum) and then divide by the #. Defined as
a balance point in a distribution of scores.
 Median-50% are above and 50% are below the score.
Defined as the middle point in a distribution.
 Most often found in nature and used as the basis for a number
Insensitive to extreme scores. of inferential statistics.
 Mode-Most frequently occurring score. May have
 Mean, median and mode are equal.
more than one mode.

Measure of Variability Inferential Statistics

 Concerned with the spread of data.  Data collection procedures that allow researchers to estimate
 Range- the difference between the highest and lowest score. how reliably they can make predictions and generalize
 Semiinterquartile range- indicates the range of the middle
findings.
50% of the scores.  Allows us to compare groups and test hypothesis.
 Standard Deviation-most stable and most useful, provides an  Answer research question in a study.
overall measurement of how much participants scores differ  http://youtu.be/lgs7d5saFFc
from the mean of the group.
 Z score-used to compare different measurements, scores are
converted to Z scores and them compared.

Inferential Statistics Inferential Statistics

 Parameter- a characteristic of a population.  2 important qualifications must be conducted to use
inferential statistics.
 Statistic- characteristic of a sample.  Sample must be representative (drawn with probability, some
form of random selection).
 Scale used must be either interval or ratio level of
 Not possible to study the whole population so we study a
sample and make predictions or statements related to our measurement.
findings.  If nonprobability sampling occurs techniques such as power
analysis are used to compensate for this.

3
Sr Dr Mona Isa FSPU UiTM Shah Alam 6/3/2016

Inferential Statistics Inferential Statistics

 Researchers are able to make objective decisions about the  Probability- the notion that in a repeated trial/study under
outcome of their study by using statistical hypothesis testing. the same conditions we would get the same results.
 Scientific hypothesis is what the researcher believes will be
the outcome of the study.  Statistical probability is based on sampling error. The
 Null hypothesis is what can actually be tested by the tendency for stastics to fluctuate from one sample to another
statistical methods. is known as sampling error.
 Inferential stats use the null hypothesis to test the validity of a
scientific hypothesis.

Type I and Type II Errors Level of Significance

 2 types of errors in statistical inference.  The probability of making a type I error.
 Type I- researcher rejects a null hypothesis when it is actually true.
 Type II- researcher accepts a null hypothesis that is actually false.  Minimum accepted level for nursing research is 0.05.
 Type I errors are considered more serious because if a researcher
declares that differences exist when none are present the potential
 “ If I conduct this study 100 times, the decision to reject the
exists for patient care to be adversely affected.
null hypothesis would be wrong 5 times out of 100”
 Type II errors occur when sample is too small.

Parametric and Nonparametric Statistics

LOS are used to determine significance.
 If wanting to assume smaller risk level will be set at 0.01.  Parametric have 3 attributes:
 Meaning researcher is willing to be wrong only once in 100 1. Estimation of at least one population parameter.
trials. 2. Require measurement on at least an interval scale.
 Decision to use alpha level 0.05 or 0.01 depends of the study 3. Involve certain assumptions about the variables being
significance. studied.
 Decreasing the risk of making a type I error increases the risk  Variable is normally distributed in the overall population.
of making a type II error.  Most researchers prefer parametric statistic when possible
because they are more powerful and more flexible.

4
Sr Dr Mona Isa FSPU UiTM Shah Alam 6/3/2016

Nonparametric Most Commonly Used Inferential Statistics

 Not based on the estimation of population parameters;  Parametric  Nonparametric
usually applied when variable measured on a nominal or  T tests whether 2 group  Chi-square- used when
ordinal scale , or distribution of scores is severely skewed. means are different. data is at the nominal level,
 ANOVA determine difference
 ANCOVA
between groups. Robust
and used with small
samples.
 Fisher’s exact probability.

Tests of Relationships
 Interested in exploring the relationship between 2 or more
variables.
 Studies would use statistics to determine the correlation or
degree of association between 2 or more variables.
 Pearson r, the sign test, the Wilcoxon matched pairs, signed
rank test and multiple regression.