C88 - L 7 - Professor Gonzales PDF

Preview

Summary

Professor Gonzales ...

Description

C 88 - L 7 ● NOIR - Nominal, Ordinal, Interval, Ratio ○ Nominal is the simplest, about categories ■ Ex. What’s your favorite color? Red, Blue, White, etc. ○ Ordinal is categories put in order ■ We know the order, but not the difference between each ○ Nominal and ordinal are noncontinuous ○ Interval is any measurement given in the form of a range ■ Ex. Rate how much you like your snack from 1 - 5 ○ Ratio ■ Has an absolute zero, there is nothing lower than zero ■ Ex. Weight, height, age ■ Normally a physical measure ■ Has ○ Ratios and integral are continuous ■ Can be expressed as a number or a fraction and still make sense ■ Ex. 144.789 cm ■ You would be able to average everyone’s scores ● Ex. On average, people gave this movie 4.6/5 stars

● ○

● ●

Pink Parscore

Internal Consistency: A Closer Look ○ When we make composite measures, we have to make sure they all go together

■

Ex.

●

● If we remove funny, we’ll have a higher internal reliability ○ Chronbach’s Alpha - We want it to be high, it’s between 0 - 1 ○ When we look at just credibility, it is unidimensional ■ When you collapse all these measures, it’s called an index Multiple Sets of Scales

●

Ex.

●

●

●

○ The green describes confidence and the grey describes expertise ○ This is why it’s a multidimensional index of credibility Coding Data ○ Often, we want to categorize data (AKA code it) ○ Pick up subthemes, which can be turned into themes ○ Usually we do this with observations or interview transcripts (non numerical) ○ Researchers may want to pick out recurring themes and put them together ○ The aim is to see emerging patterns - to help them see a bigger picture ○ Coding scheme - Rules by which to code ○ Ex. You take pictures of a certain phenomenon ■ You have a list of photo categories, (People on the phone, face to face, in business attire, etc.) ■ You can come up with a criteria on how to code your data ■ If 5 others repeated it and used your criteria, would you agree with them for where they place each photo? ■ You have to give a good coding scheme Reliability in Coded Data ○ You want to know how reliable the coding scheme is ○ You measure multiple coders choices and get a measure for intercoder reliability ○ You also want to know how consistent each coder is over multiple observations ■ If you had 20 different sets of 100 photos each time, and you repeated the task on each set, is it consistent? ○ When you measure the same coder’s choices, you get a measure for intra-coder reliability ■ How reliable is each coder compared to others? Validity of Measurement ○ Is it consistent? ■ Reliability ○ Does it match up to reality?

●

●

●

■ Validity ○ Does your measure really capture the concept you intend to be measuring? Valid Measures Should… ○ ...Be as true to what they’re measuring as we can expect them to be ○ Nothing is perfect, we try to get the most valid measures but can’t always ○ ...Have a good conceptual fit with variables in the hypothesis/RQs ○ ...Minimize potential “social desirability” bias effects ■ The tendency of survey respondents to answer questions in a manner that will be viewed favorably by others ■ Can take the form of over-reporting “good behavior” or under-reporting “bad” behavior Subjective vs. Objective Validity ○ It’s about the criteria you use to est. validity ○ Subjective Validity Ex. ■ They measure looks good on the face of it, based on your knowledge ■ According to my informed opinion, this is right ○ Objective Validity Ex. ■ I can predict the out come based on the prior studies, etc. ■ If a measure is shown to predict scored on an appropriate future measure within some acceptable margin ■ Ex. we expect SAT can predict college GPA ● If they have a relationship, then they are objectively valid Relationship between Reliability and Validity ○ Validity is harder to assess, needs more proof. Relatively is an easier formula ■ Something can be consistently wrong. Ex. A broken clock is consistently wrong, but not valid ○ Something cannot be valid but unreliable ○ You need both in science...