RM 1 - Jack Yates - Research Methods Lecture 1 PDF

Preview

Summary

Jack Yates - Research Methods Lecture 1...

Description

**= On test VARIABLE CONTROL & The Basic Experiment IV=independent variable DV=dependent variable EV=extraneous variable

Examples of experiments: ●

The effect of caffeine on RT ○ (The group got the caffeine and then you measure the reaction.)

●

The effect of the presence of a surrogate mother on exploration ○ Dependent Variable ■ Exploration ○ Independent Variable ■ Presence of surrogate mother ○ Hypothesis ■ The effect of the presence of a surrogate mother the monkey will explore more than absence of a surrogate mother.

●

The effect of birth order (organismic) on IQ EXTRANEOUS VARIABLES

● Experiments are designed to test CAUSAL relationships among variables. ● The experiment begins with the statement of a hypothesis or research question. ● The hypothesis states a CAUSAL relationship between IV and DV. ○ [IV] >[DV]

● The experiment tests this relationship by systematically varying the IV (the left side of the equation) and watching to see if the DV (the right half of the equation) also varies in step with the IV Compare to correlation. We cannot infer causation from correlation because; 1. directionality: ● does A cause B or B cause A? (traffic fatalities and economic development) 2. 3rd variable: ● perhaps C causes both A & B? (condition of the roads) How does the experiment resolve these problems? 1. the IV comes FIRST 2. The logic requires that NO EV's (3rd variable) VARY ALONG WITH THE IV. (i.e., no other 3rd variable be correlated with the IV) ● Otherwise RIVAL INTERPRETATIONS (rival hypotheses) Confounding-- when some EV varies with the IV. Internal validity --when no EV varies with the IV. ● Control preventing a variable from influencing the GENERAL outcome of a study (usually a difference between two or more groups) by preventing EV's from being CORRELATED with IV. ● Control does NOT necessarily mean to remove or eliminate a variable from the experiment How to CONTROL unwanted EVs and avoid confounding: ● HOW is each implemented, and WHY does each lead to CONTROL? ○ 1. hold constant -- set the variable to some constant value

■ Example ● How hydrated are the monkey’s getting ● If water is in an experiment, than we hold it constant ○ 2. eliminate -- set the variable to zero ■ Example ● Excluding physically disabled monkeys from the study ● Match the age of the baby monkeys ○ 3. matching -- pairs of monkeys have the same value on some EV ■ Example ● Monkeys with similar levels of energy ● Randomly assign monkeys ○ 4. balancing -- every level of the variable appears equally often in each condition of the experiment ■ Discrete variables only ● Example ○ Split by gender of the monkey’s into two groups ○ 5. within subjects - each subject serves in all conditions of the experiment (serves as own control). ■

-Ultimate in matching.

■ Counterbalance ● Order Effects ■ -Problems: practice, familiarity, fatigue, boredom -counterbalancing to control order and carryover effects

■ Example; ● Have the same 8 monkey’s participate in presence and absence of mother. ○ 6. randomization ■ http://www.ruf.rice.edu/~lane/stat_sim/sampling_dist/index.html ■ Example ● Randomly selecting groups of monkey’s by flipping a coin to participate in both sides of the experiment. Try for height*** a. Order participants from least to greatest on the EV b. take them in pairs, triplets, quads, etc. c. RANDOMLY assigned to conditions of the experiment POINT: height still varies in the groups but it CANNOT cause an average difference between the two groups If you have extra participants they can randomly assign themselves**

Counterbalanced using: Latin Square

CounterbalancedM M Latin Square Abse Pres

A

1 ent

Or de rs B

2

C

2

2 nt

Partly Present 3

1

2

3

1

controls order and carryover (practice & fatigue) effects-- time effects each condition of the IV appears in each time slot once Conditions

O r

time slots

d e r s

Latin squares for 2, 3, 4 level IVs.

Some control techniques: prevent an EV from varying:

+elimination +holding constant "cancel out" an EV in a precise way: +balancing +counterbalancing +matching +within subjects "cancel out" an EV on the average: +randomization EV's controlled do not need to be known OR measured: randomization within subjects...