Measurement and Scale Lecture PDF

Preview

Summary

MARK4000 Tuesday February 9 Measurement  Depth interview takeaways: - Product category selection - Importance of follow up questions - Get specifics - How much is enough? - We will have to do depth interviews for Assignment 2.  Measurement: Assigning numbers to objects to represent quantities of a...

Description

MARK4000 Tuesday February 9 Measurement 

Depth interview takeaways: - Product category selection - Importance of follow up questions - Get specifics - How much is enough? - We will have to do depth interviews for Assignment 2.



Measurement: Assigning numbers to objects to represent quantities of attributes. - Why do we assign numbers? - Why is it important to be able to identify what type of measurement scale we are using? TEST QUESTION: Four types: NOIR. - Ratio - Interval - Ordinal - Nominal - Higher levels of measurement have all the properties of lower levels of measurement. (Ratio contains the most amount of information)



Nominal Scale        

Ex: yes-1; no-2. All nominal has is category information. Meaning of numbers: Identification, or membership in a category Characteristics: No order, no equivalent distance, no natural zero Appropriate: Only frequencies Descriptive statistic: Percentage for each category, Mode Gender is nominal This gives you quite little information.

Ordinal Scale     

Ordinal contains the information that nominal has (category information), but it also adds something else: There is order … The numbers actually mean something. Meaning of numbers: Relative order (rank) on a variable. Characteristics: Order, but no equivalent distance nor natural zero. Appropriate: Frequencies Descriptive Statistic: And position – Median

Interval Scale  With interval, you add something to the others: The distances on the scale mean something. - Zero still doesn’t have any natural information.  Meaning of numbers: Allows relative comparison of differences (intervals sizes) on scale  Characteristics: Order, equivalent distance, no natural zero  Appropriate: Frequencies, position  Descriptive statistic: And addition/subtraction of responses between subjects, Means  If you have interval statistics or higher, you can use any of the usual statistics. - It’s kind of the minimum threshold.  Summated ratings scales aka Likert Scales like this can also be argued to be ordinal. - Why? Psychologists. - Nonetheless, in analyses, they are treated as interval. Ratio Scale  

Meaning of numbers: There is a natural, or absolute, zero; which allows comparison of absolute magnitudes. Characteristics: Order, equivalent distance, and natural zero. - The zero means the absence of whatever you’re trying to measure.

Likert scales cannot have a meaningful 0; you can always be a little less satisfied. (If negatives are impossible, a natural zero is not there). Zero has to be the absolute absence of something. Scientists use Kelvin because there is an absolute zero, which lets them make better statistical analyses with a ratio scale than without. Appropriate: Frequencies, position, addition/subtraction and multiplication/division of responses Descriptive statistic: And multiplication/division of responses between subjects, Ratios -

 

Categorical vs. Continuous Scales  

Nominal/Ordinal  Categorical Interval/Ratio  Continuous aka Scale

Remember…  It is the properties of the attribute itself that determine which levels of measurement are possible. - NOT the measurement item.  So if you use an item based on a ratio scale to measure an ordinal attitude, the level of measurement is still ordinal. Multiple Item Scales 

Necessary to measure complex, hypothetical variables (or “constructs”) - Hypothetical construct: something that exists theoretically but can’t be directly observed or measured - Requires substantial pretesting to be sure the scale is accurate (exploratory research) - Ex: Service quality (servqual), Brand equity o Brand equity refers to the added value that a brand brings to a product or service. o Brand equity is usually considered to have three dimensions:  Broad familiarity….

Accuracy of Multiple Item Scales  Accuracy is assessed through tests of validity and reliability Reliability vs. Validity  Relationship between validity & reliability: - If a scale is perfectly valid, it is also perfectly reliable - If a scale is unreliable, it cannot be valid. - If a scale is perfectly reliable, it may or may not be perfectly valid. o Reliability is a “necessary but not sufficient” condition for validity  That is, we may measure something the same way every time, but are we measuring the right thing?...