Canonical Correlation

Is use to determine the relationship between 2 or more Predictors Variables and 2 or more Criterion Variables  

Fundamental Attribution Error

Some cultures tend to favor situational factors for the behaviors of others

Research on Children Illness sugest that young children?

Young children that are given more complete information about illness, diagnosis, prognosis and treatment have better psychosocial outcome compare to children who are given limitted information

Information need to be given in developmental appropriate manner           Bearison 1994

ANCOVA

Reduce error variance that is attribute to control variables

Selection Test in a Criterion Validity Study. The investigator need to?  

^ Increase test cutoff score to

Decrease false positives

Criterion Related Validity Coefficients for a selection test on Females/Males  where Female score is statistically significant but not males

This is due to differential validity  

T-score of 65

= 1.5 standars deviation

Reliability is below .65

Investigator need to use subjects who are more diverse

Reliability should be .80

If a test on one area correlate highly with a test on another area

example cinysm correlate with anger

this is due to?

Low Construct Validity

The coding categories of an ^Increase inter-rater reliability of a behavior observation scale

The coding categories need to be mutually exclusive

Sampling Errors

Tendency of sample statistic to vary from one another due to the effects of randon errors

Ramdon selection of sample from population is an assumption of

Parametric and nonparametric test

Variable Ratio

Schedule to reinforce a behavior every X amount of times it occures

A research professor hires a student who does mayority of the work for a book. The professor should list him as a 

 Co-Author

Variable Ratio

Maximize resistance

Fixed Ratio

Extinguish Behaviors

Multivariate technique known as path analysis is to evaluate?

The viability of a casual model for a set of variables

Internal validity

A study has good internal validity if the effect of the independent variable on a dependent variable are in fact due the the independent variable (IV).

Factor that can threaten internal validity:

• Maturation -biological/psychological changes within subjects during curse of study due to time and affect the DV sytemically
• History affect when an external factor affect the status of subjects on the DV.

External validity

Good external validity if it allow researcher to generalize to other people, conditions or contexts. Limited by:

• Low internal validity- cant determine if the IV has an effect on the DV → then cant generalize

Single subject design

Includes:

• A-baseline or no treatment phase
• B-treatment phase
• DV-is measure at regular intervals during each phase

ABAB design

Icludes four phases:

• DV is measure initial baseline,
• While tx is administered
• After finish, withdrawal or stop of tx
• While tx administer a 2nd time

Multiple baseline design

Involves:

• apply tx across different baselines-can be different subjects, behaviors or settings. tx don't have to be removed
• example apply same tx to nail-bitting, then to screaming behavior and then to failure to make eye contact 

Group design- 3 types  

• Between groups
• Within subjects and time-Series design
• Mixed

• Between subject-each group is exposed to different level of the IV so that comparisons are made between groups.
• Within subject-each subject is exposed to all levels of the IV at dif times (Level 1, Level 2 and so far). Time Series- Measure DV at regular intervals several times before and after the IV is admin to subjects
• Mixe-more than 1 IV with one IV been in between-groups while the other are within subjects  

Factorial Design

Any group design that inclusdes 2 or more IV

Provides data on the main effects of each IV as well as on interaction bwteen IVs 

Experimental Research

Has at least 1 IV and 1 DV

Involve testing hypotheses about the relationship between those variables

• True-experimental → the investigator has control over the variables and condition of study and can randomly assig subjects to diff levels/tx groups  of the IV. this prove that the IV has effect on the DV.
• Quasi-experimental- the investigator cannot randomly assign subjects to the dif levels bc he need to use intact group (group A in clinic A-group B in clinic B) or bc IV is organismic (self esteen or intelligence)

Inferential Statistic - base on probability so decisions made may or may not be correct. this is independent of rejecting the Null Hypothesis 

Two type of incorrect decisions:

• Type I error
• Type II error

• Type I Error- a true Null Hypothesis is rejected. The prob of Type I error is equal to alpha/the level of significance. alpha set before the stydy example - .05 margin of error  
• Type II Error- a false Null Hypothesis is retain . the prob of making this error is due to beta/not control by the investigator     

Power

Power is desirable

Ability to reject false Null Hypothesis.

Power is maximize if

↑ Alpha ↑ sample siza ↑ effect of IV

↓ error, use one-tail test if appropriate, and use parametric test to analize the data

Chi-Square Test

Use with nominal/categorical data

Certain assumptions needs to be met - independence of observation → no relationship between subject in one category and subject in the another category. Violated if one subject is listed in more that one category 

ANOVA- is use with date representing an internal or ratio scale

Trend analisis - analisi of variance use when the IV is quantitative ► investigator can determine if linear or non-linear relationship between IV and DV

example - antidepresant given at dif dosis ► the higher the dosis = more positive efftects

Trend Analisis - indicate if any trend is significant

MANOVA- multivariate analisis of variance

Has 1 or more IV and 2 or more DV

All DV are measure on interval or ratio scale

Assess the effect of IV on all DVs

Reduce experimental/Type I error  

Pearson r - correlation coefficient

use when both variables are measure on a continuos interval or ratio scale

Use when linear relationship between variables

Point Biserial - correlation coeficient

Use when one Variable is continues and the other is a true or artificial dichotomy

Use when a linera relationship between variables

ETA - correlation coeficient

Use when both variables are on a continous scale and their relationship is non-linear

To interpret correlations

square - to produce coefficient od determination ► measure of share variability

R- multiple correlation Coefficient ►correlation between 3 or more variables can be squared to measure share variability  

Multiple Regression 

Each predictor should have high correlation with criterions but low correlation with other predictors

Multicollinearity- predictors that correlate highly with each other ► not desirable bc it provide redundant and not unique information about the criterion

Cohen's d, r and eta

Most comun method for calculating an effect size

Example- study shows that x treatment's effect is statistically significant

Cohen's d: measure the diference between two groups  (experimental and control group)

     mean of group a

   - mean of group b

   / standard deviation for both groups

Small effect size = 0.2

Medium effect size = 0.5

Large effect size = 0.8

Cohen's continue...

r square (r2) - % on variance in the outcome accounted for variance in treatment    

eta square (n2) - % on variance in the outcome accounted for variance in treatment