PowerPoint presentation to accompany Research Design Explained 6th edition ; ©2007 Mark Mitchell & Janina Jolley Chapter 7 Introduction to Descriptive Methods
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Overview l Uses and Limitations of Descriptive Methods l Why We Need Science to Describe Behavior l Sources of Data l Describing Data From Correlational Studies l Making Inferences From Correlational Data
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Uses and Limitations of Descriptive Methods l Descriptive Research and Causality –Can’t test causal hypotheses Can show that A and B are related, but can’t show whether A causes B, B causes A, or both are effects of some other factor –May suggest causal hypotheses l Description for Description’s Sake l Description for Prediction’s Sake
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Why We Need Science to Describe Behavior l We need objective scientific measurement to overcome the human tendency toward bias l We need systematic, scientific record-keeping because memory is selective l We need objective ways to determine if variables are related because humans don’t innately compute correlation coefficients l We need scientific methods to meet both criteria necessary for making accurate generalizations (1) obtaining a representative sample and (2) making statistical inferences from that sample data
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Sources of Correlational Data l Ex post facto data l Archival Data l Observation l Tests
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Ex Post Facto Data l You may have collected it while doing an experiment l External validity: Depends on sample l Construct validity: Depends on validity of measures l Internal validity: None
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Archival Data l Some has been collected and coded by others l Some is part of a public record (transcripts, web sites, videotapes, personal ads, etc.) l To code uncoded data, you will probably use content analysis
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Archival Data (cont) l External validity: May be good: Large sample possible. l Construct validity: –May be good because measures can be nonreactive. – However, if data have been coded by others, their poor coding and/or instrumentation bias may hurt validity. If data were uncoded, validity can’t be better than the validity of your content analysis. l Internal validity: None
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Observation l Lab observation l Naturalistic observation l Participant observation
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Conclusions: Validity of Observation l External validity: Depends on sample l Construct validity: May be damaged by –Observer’s presence changing participants’ behavior –Observer not accurately recording behavior l Internal validity: None
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Tests l External validity: Depends on sample l Construct validity: Usually good l Internal validity: None
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Describing Data From Correlational Studies l Graphing Data Graph of a strong positive correlation** Graph of a strong negative correlation** l Correlation Coefficients Graph of a correlation coefficient** Graph of a correlation coefficient** Graph of a 0.00 correlation coefficient**
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Perfect Positive Correlation
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Strong Positive Correlation
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Perfect Negative Correlation
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Strong Negative Correlation
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Zero Correlation
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Mathematical Notes About Correlation Coefficients 1. Sign indicates direction of relationship, but not strength 2. Absolute value indicates strength of relationship (farther from zero, the stronger the relationship) 3. Squaring the Pearson r gives you a measure of the strength of the relationship-- the coefficient of determination, which ranges from 0-1
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Mathematical Notes (cont.) 4. The type of correlation coefficient you should compute depends on the type of data you have –If both variables are interval or ratio, Pearson r –If both variables nominal, phi coefficient –For more details, see Table 7-4
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Making Inferences From Correlational Data l Analyses Based on Correlational Coefficients l Analyses Not Involving Correlation Coefficients l Interpreting Significant Results l Interpreting Null Results
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Are the two variables related in the population? l Need random sample of population l Statistical test to determine if the variables are related –Several tests to choose from –All will be more likely to say that the variables are related if l The correlation coefficient is large l Sample size is large
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Tests Used to Determine If Variables Are Related l t test** l ANOVA** l Test to see if the correlation coefficient is significantly different from zero**
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley T-test –If one of the variables has only two values (gender), t test works well –If both variables are continuous, have to l Use median split to create two groups l Live with concerns that the median split has lost you power
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley ANOVA –More power than t (if divide participants into more than two groups), but less than directly testing to see if the correlation is significantly different from zero –Convenient, familiar way to l Look for nonlinear trends l Look for interactions
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Testing to See If the Correlation Coefficient Is Significantly Different from Zero –Simple, direct test –If both variables are continuous, it is more powerful than both the t and ANOVA test
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Cautions about Significant Results l Don’t allow cause-effect statements l May represent Type 1 errors, especially if –Numerous tests were done and –No corrections were made for the number of tests done
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Cautions about Null Results Null results may be due to l Not enough participants scores l Insensitive measure(s) l Nonlinear relationship l Restriction of range l Using a t test rather than the more powerful test
Research Design Explained, 6th Edition; ©2007 Mark Mitchell & Janina Jolley Concluding Remarks l You now know the basics of descriptive research l However, to learn about the most commonly used descriptive method (the survey), you need to read Chapter 8