Experimental

Challenge of Between Subjects Design:

...

Things to NOT do with a graph:

1. Have different scales on similar graphs 2. 3-D graphs 3. Graphs for items that can easily be written (in less space) 4. Overly complicated graphs

We aren't good at probabilistic reasoning:

1. Insufficient use of probabilistic Information (Base-rate neglect) 2. Failure to take sample size into account 3. Failure to understand randomness

The True Experiment: 3 characteristics

1. Random Assignment 2. At least two levels of an independent variable. 3. have to have major controls for the threats of internal validity.

Theory driven research vs. Direct application

Although some research is designed to predict events directly in a specific environmental situation, much scientific research is basic research designed to test theory.

Within-subject design

Assumes random assignment. This has a 2 group design. Removes error brought about by individual differences. Use a paired sample T-test.

Relationship between what the error bars represent and B/W subject variance

B/W subject variance: variance that is due to the independent variable, the difference among the three groups. For example the difference between a person's score in group one and a person's score in group two would represent explained variance.Error bars account for the variance of the mean score that the IV impacted.

Answers to college sophomore problem 3

College students are extremely diverse. Replication of findings ensures a large degree of geographic generality and, to a lesser extent, generality across socioeconomic factors, family variables, and early educational experience.

Advantages of BSD

Each individual score is independent of the other scores. Participant's scores are not influenced by practice or experience, fatigue or boredom, contrast effects.

Between Subjects Design:

Each participant participates in one and only one group. The results from each group are then compared to each other to examine differences, and thus, effect of the IV. Each participant only has one score.

GF: We fail to realize that rare events should happen with some frequency, based on chance and nothing more

GF is not restricted to games of chance. Operates in any domain in which chance plays a role. The probability of having a boy is exactly the same after having two girls as it was int he beginning.

Multi-level BSD

Have a single factor that has a higher rank than 2. Has more than two levels. A one way ANOVA is used. (Mean of G1=Mean of G2=Mean of G3) Then you would use a post hoc T-test.

Threats to internal validity 1

History: there could be something happening between the two events (tests/measurements) of the experiment that can effect experiment and results. This is solved with control groups.

We try to explain random events

Illusory correlation: when people believe that two types of events should commonly occur together, they tend to think that they are seeing co-occurrences with great frequency, even when the events are occurring randomly. They do not co-occur more frequently than other combination of events.

What can the results of a true experiment tell us that we can't otherwise know?

It is randomization that makes true experiments so strong in internal validity and typically allows us to make relatively strong influences about causality

Disadvantages of BSD

Large number of participants and between subject variance, Individual differences (differences that range from one participant to another) can produce high variability in the scores and can become confounding variables.

What do error bars represent?

Measures of variance. Locate where the real population mean lies. They tell you the distribution of sample means of two groups.

Threats to internal validity 6

Mortality: Lose track of people, drop out, stop coming, etc.

Connectivity principle

New theory must explain old facts. It has to explain what came before and connect it to the past. The connectivity principle stresses that new theories and ideas must not only present new information, but must also account for old information. This principle requires that scientific progress is a cumulative process.

Insufficient use of probabilistic information: Base-rate neglect

Not looking at all of the statistics and only looking at one variable. Improper framing. Tendency to give insufficient weight to probabilistic information Example: When testing for HIV, medical personnel assumes that since people tested positive for HIV that they actually had the virus, but only 2 % actually did

Statistics are important:

Once the relevant variables have been determined and we want to use them to predict behavior, measuring them and using a statistical equation is the best procedure.

Factorial Design interactions:

Parallel lines=interaction non parallel lines=no interaction Combine in ways that give you different data than if you looked at them individually.

2 group BSD

Participants have been randomly assigned to the two levels of the independent variable. Use an independent sample T-test. (mean of G1= mean of G2)

We do not generate random strings

People routinely underestimate the likelihood of runs and patterns in a random sequence. This causes people to not be able to generate sequences or runs when trying to do so. They usually will have too few of each.

What is the function of a graph:

Present information to readers in a meaningful and manageable format. Graphs are an excellent way to display information visually. They also support the information and results you have presented.

Internal Validity vs. External Validity

Random sampling --> External validity Random assignment --> internal validity This is a tradeoff between real world applications and extreme control (internal validity) in a lab.

Gradual synthesis model

Scientists are able to make gradual theories about human behavior after looking at large amount of evidence to support one theory.

Why is random assignment important?

Scientists can rule out alternative explanations of data patterns that depend on the particular characteristics of the subjects. Ensures that the people in the conditions compared are roughly equal on all variables because as the sample size increases, RA tends to balance out chance factors. Rules out systematic bias in how the subjects are assigned to the 2 groups.

Why can laboratory settings be good?

Scientists deliberately set up conditions that are unlike those that occur naturally because this is the only way to separate the many inherently correlated variables that determine events int he world. The lab allows for more precise control.

Threats to internal validity 5

Selection: might have differences of subjects in the different groups. Solved with random assignment

Clinical Prediction doesn't work because

The experience of psychological practitioners allows them to go beyond the aggregate relationships that have been uncovered by research.

College sophomore problem:

The worry that, because college sophomore are the subjects in an extremely large number of psychological investigations, the generality of the results is in question.

What do control groups tell us?

They basically tell us what we need to know. They are sometimes more effective than the experiment group. EX: treatment/no treatment group (placebo effect) In this example, the control group allows the researcher to study the effects of the treatment thoroughly.

Goal of Between Subjects Design

To determine whether differences exist between two or more treatment conditions.

Factorial Design

Two independent variables with 2 levels (2x2). You need fewer participants. This design shows an interaction between variables. When two variables act in a way that is different than what there sum is. Use an ANOVA for any factorial design. Main effects are just like T-tests between groups.

Failure to take sample size into account:

We hold generalized beliefs about larger populations. Other things being equal, a larger sample size always more accurately estimates a population value.

Scientific Consensus

We recognize that all experiments and everything has flaws but we learn something when everything can come to a consensus and to one conclusion. Most science isn't a breakthrough, it is incremental, brick by brick.

Failure to understand randomness

We tend to focus on alternation and not enough long strings of #s. We fail to realize that rare events happen by chance. We tend to have problems seeing false links between past and future.

Why is counter-balancing important?

When diff aspects of an experiment are sequentially presented to participants, it's important to consider the order of presentation because the sequence of events could influence the dependent variable and it needs to be controlled. Ex: When we use medication and supportive therapy with depression patients then be sure that sequence of experiment first demands medication and then supportive therapy

converging evidence

When evidence from a wide range of experiments, each flawed in a some what different way or carried out with techniques of differing strengths and weaknesses, points in a similar direction, then the evidence has converged. Urges us to base conclusions on data that arise form a number of slightly different experimental sources. Gradual, emphasizes the practice of examining the related evidence in order to come to a conclusion.

Psychological causes are all probabilistic

When the issues are psychological, people tend to forget the fundamental principle that knowledge does not have to be certain to be useful--that even though individual cases cannot be predicted, the ability to forecast group trends accurately is often very informative.

Gambler's Fallacy:

belief that events in the past and future are related even though they are independent of each other. Example: You get heads five times in a row when flipping a coin, so you think that this has an effect on what will happen in the future when it is just a chance event

Threats to internal validity 4

changes in instrumentation: changing measurement tools, room, etc. It is basically anything that you change. This can be solved with control groups.

Importance of converging evidence

conclusions in psychology are often based ont he principle of converging evidence. Data that support a given theory usually rule out only a small set of alternative explanations causing conclusions to be possible only after data from a very large number of studies have been collected and compared.

Cognitive Illusions: (Go with base-rate neglect)

even when people know the correct answer, they may be drawn to an incorrect conclusion by the structure of the problem.

Internal Validity:

goal is to pinpoint affects of dependent variable on independent variables by ruling out alternatives.

direct application/applied research

goal is to relate the results of the study directly to a particular situation. When the nature of the question is direct, questions of the randomness of the sample and the representativeness of the conditions are important because the findings of the study are going to be applied directly.

Answers to College Sophomore Problem 1

it does not invalidate results but simply calls for more findings that will allow assessment of the theory's generality. College students are people too and if it doesn't work for them than it's different for all people.

Answers to College Sophomore Problem 2

it is simply not an issue because the processes investigated are so basic that virtually no one would believe that their fundamental organization depends on the demographics of the subject sample. The processes are very basic: it applies to humans, not just college sophomores.

Threats to internal validity 2

maturation: aging or passage of time. This can be solved with control groups.

Taxi Cab example:

p(blue)=.95, p(green)=.05 p(witness correct)=.9, p(wit. not correct)=.1 "probability distribution"

Actuarial prediction

predictions based on group trends derived from statistical records. They predict the same outcome for all individuals sharing a certain characteristic. We get more accurate predictions. Public knowledge.

Progress despite flaws

progress only when we explain more or explain the same amount more simply. Research on a particular problem often proceeds from weaker methods to ones that allow more powerful conclusions to be drawn.

Threats to internal validity 3

repeated testing: causes particpants to act and respond differently in future tests (realize something they are doing and then stop it) EX: Hair twirling. This can be solved with control groups.

Theory driven research

seeks to test theories of psychological processes rather than to generalize the findings to a particular real world situation. "basic research".

Clinical prediction (case prediction)

some subgroups of clinical psychological practitioners claim to be able to go beyond group predictions and to make accurate predictions of the outcomes of particular individuals. *doesn't work*

Random Assignment

subjects themselves do not determine which experimental condition they will be in but, instead, are randomly assigned to an experimental group. All participants have the same likelihood of getting into each group. Key to develop conclusions

Great leap model

supports the idea that scientific problems and issues are explained/solved when an experiment overturns all previous theories and completely explains the concept of interest. It either explains previous theories in a revolutionary way, or it completely dismisses them. Violate the Principle of Connectivity, which states that new theories should connect with old theories by supporting and building on previously established facts.

Control Groups

the group treated just like the experimental group except for the absence of a critical factor. Rule out possible alternative explanations. You can never get it perfect every time and can always rule out more possibilities.

We try to predict the unpredictable

we are reluctant to acknowledge the role of chance when trying to explain outcomes in the world. We must accept errors in order to make fewer errors overall. Statistical information should never be set aside when one is predicting behavior.

We see connections that aren't there

when people have prior belief that two variables are connected, they tend to see that connection even in data in which the two variables are totally unconnected.