What is a One-shot Case Study?

(Pre-experimental Design)

A single group is exposed to a treatment and a dependent measure is assessed.

-NO random assignment of people or conditions.

You give a pie to someone (or a group) and measure his or her glucose level afterwards.

Qualitative and Testimonials fall into this kind of design.

What is a One-Group Pretest-Posttest Design?

(Pre-experimental Design)

This is like the one-shot study except there is a pretest.

Your analysis compares performance on the post-test with the pre-test.

For example, you measure a pereson's glucose, give her a pie, and measure the glucose again.

Extraneous variables are NOT controlled.

What is a Post-test Only Design?

(Pre-experimental Design)

In this type of design, there is an experimental and control group, but they are only compared on the dependent measure AFTER the expimerntal treatment.

There IS AN EXPERIMENTAL CONTROL GROUP.

What is a Pre-test/Post-test Experiment?

(True-experimental Design)

Pre-test important because it allows them to compare drop-outs from the experiment with those who stay in.

Pre-test scores - you can exclude outliers

Pre-tests can have their down-side. (add a lot of time to the study and require more work of the research to administer and score, sensitize the participant to the kind of skills/point of view)

You must decide whether you will present the subject with all levels of the independent variable (Called repeated measures) or whether you will have separate groups of for each level of the dependent variable.

Repeated Measures - fewer participants, uses a person as their own control group, fatigues participants, sensitivzes person with the types of problems they are dealing with.

Morality refers to the loss of participants from a study.

It is important to use a pre-test design because people who drop out of the study might be different on something important from the participants who remain in the study.

A control and an experimental group

(participants randomly assigned to groups)

Pre-test and Post-test

Dependent variables are measured for both groups after the exposure of the experimental to the independent variable.

What are Matched Groups?

(Complex Design)

In complex designs, the participants (while randomly assigned) in each group are generally matched in relevant characteristics with the participants in other groups.

The groups are alike on relevant factors or on ones that might remotely be relevant (such as gender, muscle mass, or age)

What are Elaborate Independent Variables?

(Complex Design)

These designs either have more indpendent variables or more levels of the one that they have.

-Researchers want more information about the IV & DV, want to be able to detect curvilinear relationships and effect of many treatments.

Example, think of different levels of difficulty, might also have gender (see chart)

What are Multiple Dependent Variables ?

(Complex Design)

Example, if we were interested in the side effects of presentation modality (how information is presented on a computer screen - as a graph, diagram,cartoon etc), we would want to test people in many ways.

-Asking them in different ways!

a) the accuracy of the student b) the speed of the student c) the ability of the student to make inferences from what he/she has learned

Multiple dependent measures.

It is called this because in many cases it is similar to a standard experimental design. Its basic limitation is determined by the way it assigns participants to groups.

Difficult to claim causality.

Participants NOT randomly assigned to groups (often matched, or the groups are conveniently assigned.

Don't always possess a high level of internal validity

What is a one-group posttest-only  design?

(Quasi Experimental)

Participants ---> Sit next to stranger----> Measure time until stranger leaves

The missing piece of this design is that there are no control groups and the "strangers" are not random volunteers and may have all kinds of special characteristics.

Roughly equivalent to the one-shot design except in this case we're observing something occurring and in a typical one shot we are doing something to someone and testing the results

What are Non-equivalent group designs?

(Quasi Experimental)

It usually has a pre-test and post-test but not random assignment of participants.

Participants -- Training program ---- Smoking Measured

Participants --- no training program --- smoking measured

Natural occurring groups - they are selected by nature, not the researcher.

-Comparing achievements of first-born children with that of later-born children

-Comparing student performance at schools, one of which has a lower student-teacher ratio.

-IQ and weight.

If people are assigned to groups non-randomly: say perhaps based on their scores on the pretest; high spatial ability people in one group low spatial ability in another.

This may result in regression to the mean so that the high scores, if their scores change it will go lower (toward the population mean) and the low scorers may tend to score higher at the end of the study (again towards the meant)

This is the type of design just described under "statistical regression"

Deliberately want to try and make low scoring people equivalent with high scoring people.

The football example ( good readers random story, poor readers football)

Formulating a Hypothesis

Selecting a design for the study

Selecting the statistical test to analyze the findings and see if the hypothesis is correct.

This is a scale where the categories (or numbers) are used to classify people.

No less than/ greater than

Schizo, paranoi, manic-depression etc

NONPARAMETRIC

This scale occurs when the categories have a relation to one another.

Take the categories from earlier (Schizo, paranoia, etc)

If these cateogires can be arranged into degrees of severity (One is more severe than the next) then it forms an ordinal scale (ordered)

FORM A RANK ALONG A CONTINUUM

NONPARAMETRIC

This is an interval scale with a true zero so that if you change the scale labels, that ratio stays the same (lbs to grams)

(Generally physical measurements of weight or time measures such as duration or reaction time)

PARAMETRIC

When the categories have a relation expressed numerically (Example, the degrees of severity)

-Not just that one cateogry has more of something than another category, it consitutes an interval scale.

In a measurement of anxiety 10-11 would be the same as 50-51

PARAMETRIC

Designed to compare observed frequencies within groups to their expected frequences.

The bigger the difference between expected and observed, the bigger the chi-square, the more likely it is that the observed frequency did not come from the population on which the null hypothesis is based.

A survey is a form of stylized interview with precise questions and limited answers that can be collated numerically.

Summary results of a survey in terms of qualitative data (typically medians, but maybe means)

P

When we obtain a percentage in favor of a particular candidate, we want to know how reliable the number is.

The margin of error is the variability around an obtained number that you would expect.

The population is divided into strata followed by random sampling from each stratum

-Area probability sampling is when the strata consist of geographic areas.

This type of statistics allows researchers to make inferences about the true difference in the population on the basis of the sample data and gives the probability that the difference between means reflects random error rather than a real difference.

However, if the groups are initially equivalent, then any differences in the dependent variable must be due to the effect of the independent variable.

Statistical significance assesses the probability that results could be due to chance rather than the hypothesized cause.

For example, can observed difference (in means) be accounted for JUST by chance? Where chance is everything not accounted for in your manipulation.

How likely it is that we have a true or real difference between groups.

Used with ordinal data, assesses the degree of relationship between two variables.

The example of the two Judges ranking the four essays. If their rankings of the four essays are positively correlated, then you would say that they're using the same standards in grading them.

-Used when you have at least ordinal data and need to do a comparison of positions in that ordering. Most often used when data do not meet the standards of normality (normal distribution), homoscedasticity (equal variability in the two distrubtion) or linearity (increasing in scores) If they meet these requirements, then the appropriate correlation would be basic, simple correlation technique.

A correlation is a single number that describes the degrees of relationship between two variables.

(Need Interval data)

This type of correlation invlolves two or more independent variables and one dependent one.

A's and Self Esteem Example.

First it tells you which independent variable is/are more important in the prediction of the dependent variables.

Second, it tells you how much of the variability in the dependent measure is explained by one or more of the independent variables

Finally, regression analysis provides the basis for an equation that relates the one or more expresses the relationship between two (more) variables algebraically.

Beneficence  - There should be a benefit for the researcher (cost/benefit analysis) to society and/or the individual participant

Autonomy - The research should treat the participant with respect

Justice - Says there should be equity in treatment and inclusion/exclusion of people within the study.

Exempt research in which there is no risk or harm

Minimal Risk Research when the risk of harm is no greater than risk encountered in daily life or routine physical or psychological tests.

Harm

Physical harm/ stress/ loss of privacy and confidentiality