COMM 301:Empirical Research in Communication Kwan M Lee Lect3_1
Announcements • Office hours • Initial survey with email hand in to TAs • Class Activities • Unresolved problems Ask TA or contact Professor
Building good measures: Levels of measurement • Things to know by the end of this lecture: • What is measurement? • What are the levels of measurement? • What is the Likert scale? • Why do we care about levels of measurement? • How levels of measurement affects statistical tests?
Measurement; Numerals; Numbers • Measurement: assignment of numerals to objects according to rules • Numerals: symbols (label; category) • Numbers: symbols with quantitative meaning. • Levels of measurement • Nominal • Ordinal • Interval • Ratio
Nominal level of measurement • Nominal level of measurement assign a numeral to a variable, without quantifying the numeral • i.e., name • Example: biological sex female = 0 male = 1 Note, no sense of quantity: “1” here does not mean more than “0”
Nominal level of measurement • Criteria for nominal level of measurement • Mutually exclusive • Every observation will fit into one and only one of the categories • Exhaustive • Every observation have one category in which it may be mapped • What pet do you like • Dog; Cat; Fish ……………… Others • Equivalent • All categories should be about the same attribute of variable being mapped. • News; Drama; Comedy……….; 30 min; 1 hour
Nominal level of measurement: Examples • Party affiliation • Dummy variable • Rep 1; Dem 2; Green 3; Independent 4; Other 5 • No rank order above!
Ordinal level of measurement • Ordinal level of measurement retains the criteria for nominal level • mutually exclusive, exhaustive, equivalent • Plus, the numerals are rank ordered from low to high • But, there is no assumption of equal spacing between the ranks
Ordinal level of measurement • Example (ordinal level) • Socio-economic status • Low; Middle; High • Top 3 social issues in the USA • Note unequal distances in above examples
Interval level of measurement • Interval level of measurement retains attributes of the ordinal level: • mutually exclusive, exhaustive, equivalent • rank ordered • Plus, that the spacing between the ranks are equal • And, there is an arbitrary zero point
Building good measures: Interval level of measurement • Example (interval level) • Fahrenheit scale • equal spacing: a change from 60 to 70 is the same as the change from 90 to 100 • arbitrary zero: temperature can fall below 0, 0 does not indicate lack of temperature • IQ • Note: the distinction b/w ordinal and interval scales are fuzzy in real life
Interval level of measurement • Implications: • we use numbers at the interval level • we can use mathematics to work with the numbers to get some meaningful results • Scales using interval level Likert scale Semantic differential scale • In Social Science Research, ordinal scales are usually used as if they were interval. • So in real research, the distinction b/w ordinal and interval is not that important.
Interval level of measurement; Examples • Likert scale • set of statements assessing (dis)agreement with a 5- or 7-point scale (usually) • Example: • Kwan is a good teacher. • Strongly disagree 1 • Disagree 2 • Neutral 3 • Agree 4 • Strongly agree 5 • Kwan is well organized • Kwan is not funny • Note: reverse coding • Sum up the response values for all statements, to get a measure of Kwan’s teaching ability
Ratio level of measurement • Ratio level of measurement retains attributes of the nominal level: • mutually exclusive, exhaustive, equivalent • rank ordered • equal spacing • But, instead of an arbitrary 0, it has a true 0 • True 0 indicates absence of the phenomenon in question
Building good measures: Ratio level of measurement • Example (ratio level) • income • age • News recall measure based on 10 items • …
Building good measures: Why do we care about the levels? • Levels of measurement determine what sort of statistical analysis you can do later. • Nominal, ordinal levels use a type of statistics • Interval, ratio levels use another type • An important point in the research process, need to think through.
Building good measures: Why do we care about the levels? • Generally, try to use higher level of measurement if possible. • Example: Age • Age measured in ratio (number of years): 21, 33, 46, 87 • Age measured as ordinal categories • Teenager • Young adult • Middle age adult • Elderly senior
Levels of measurement and an introduction to statistical tests • Statistical tests determines more precisely if there are relationships between variables. • In this course, we will focus on three statistical tests: • Non-parametric statistics: for nominal and ordinal • chi-square test • Parametric statistics: for interval and ratio • t-test • Correlation • Don’t worry about the next slides on the details of each stat tests. We will learn them one by one later in the semester.
Introduction to statistical tests Chi-square test • Chi-square test • There are several variations of the chi-square test. • In this course, we will focus on one particular variation: comparing frequencies of two or more distinct groups. • E.g., composition of a class (Sophomores; Juniors; Seniors) • Independent variable is group membership, measured at the nominal or ordinal level. • Dependent variables are observed frequencies of each nominal or ordinal category. • Examples ...
Introduction to statistical tests T-test • T-test • There are three forms of t-tests. • The one we will use the most is the independent samples t-test. • The independent samples t-test compares average (mean) score of one group with the average score of another group. • For example, the average GPA of communication majors is compared to the average GPA of engineering majors.
Introduction to statistical tests Independent samples t-test • Independent samples t-test • Independent variable is group membership, measured at the nominal or ordinal level. • Examples … • Dependent variable is measured at the interval or ratio level. • Examples … • Particularly appropriate for experimental designs, usable for surveys also. • Examples …
Introduction to statistical tests Correlation • Correlation • Correlation tests if there is a linear relationship between two variables. • Examples ... • Independent variable is measured at the ordinal, interval, or ratio level. • Dependent variable is measured at the interval, or ratio level.