1. Observational Research

2. Observational Research
Naturalistic Observation:
Unobtrusive observation (avoid the Hawthorne Effect)
Habituation
Indirect measures
Count the results of behavior, use a survey
Disadvantages: time & money
Advantages: ecological validity

3. Observational methods
EEthnography: researcher is immersed in the behavioral or social system being studied. Often used by anthropologists (skip pages )
Decide whether to be a participant or nonparticipant and overtly or covertly
Reactivity might be a problem if overt and ethical issues might surface if using covert method is used
Issues related to gaining access to a setting or group might present

4. Observational Methods
Case histories require that you study a single or just a few cases
Case studies are particularly useful when the goals is behavioral change or when organizations are studied (e.g., learning/education/ and industrial/organizational settings)

5. Observational Methods
Archival research involves studying existing records such as historical accounts, police records, published articles, or media
Requires a specific and refined hypothesis
Might consider how you will gain access to the data, the completeness of the record (do you need more than one source)

6. Observational Methods
Content analysis: involves analyzing verbal written or spoken record for the occurrence of specific categories of events, items, or behavior.
Some overlap with archival research; some people define content analysis as pertaining specifically to language while others have a broader definition.
Might include conversations, books, movies, blogs, etc.
Successful content analyses require that researchers be objective and systematic, and have clear operational definition and/or coding schemes.
Consider issues related to sampling (avoid a biased sample) and observer bias (more than one coder)

7. Behavioral Categories & Coding Schemas
Behavioral categories operationally define what behaviors are coded during the observation period
Clearly defined hypotheses
To develop categories you could make preliminary observations, conduct a lit review, or be very specific about your research goals and hypotheses
help refine your categorical codes (the behaviors that you will record)

8. Quantifying Behaviors
Frequency method: record the number of times a behavior occurs
Duration method: record how long the behavior lasts.
Interval method: divide the observation period into time intervals, record the number of times the behavior occurs within each time interval (e.g., verbal exclusion during 2 minute time periods)

9. Recording single events vs. behavioral sequences
Behavioral sequence can be thought of
ABC’s of the behavior: antecedent, behavior, consequence
Antecedent only or consequence only

10. Consider sampling or recording complex behaviors
Time sampling
Individual sampling
Event sampling: observe only one behavior
Record behaviors code later by watching video repeatedly

11. Reliability Interrater reliability: involves using multiple coders
Ensures that observers are accurate
Allows for replication
Allows coders to detect and correct any discrepancies

12. Methods of determining reliability
Percent agreement ((total # agreement/total # observation) x 100)); >70% is acceptable
Cohen’s Kappa κ = Po – Pc/1– Pc
Used for categorical or dichotomous data
where P o = observed proportion of actual agreement, and Pc = proportion of expected agreement
Pearson’s r
can be used with continuous data but it might produce a significant correlation if disagreements are numerous (as long as the magnitudes increase or decrease in a similar fashion)
Intraclass correlation coefficient (ICC)
(rI) should be used for continuous data. This method uses an ANOVA approach (means squares within and between subjects

13. Sources of bias
Observer bias: when being aware of the hypothesis influences coding. Can use blind observers
Observers interpret rather than record behavior

14. Chi square A non parametric test
chi-square & fisher’s exact test is distribution free and relies only on frequencies
tests can only be used under certain circumstances
chi-squares: dichotomous or categorical data
fisher’s exact: 2 by 2 table or two dichotomous variables.

15. Chi-square
To calculate χ2 determine the frequency of each cell if no differences existed (frequency expected, (ƒe) and then compare this to the actual or observed frequencies (ƒo).
The greater the difference between expected and observed frequencies the more likely it is that differences exist.
χ2 = ∑ (ƒo – ƒe)2/ ƒe
Compare χ2 observed to χ2 critical in the χ2 sampling distribution

16. Chi-square To report your findings:
χ2(df, N = #) = statistic value, p-value
χ2(1, N = 90) = 0.89, p = .35
Where df = (r-1) x (c-1)