Experiments (ST 311 (3))

observational studies and experiments

The two most common ways to collect data are …

Experiment

The process of applying some treatment and then observing its effects

compares

An experiment almost always … two (or more) groups. This typically, this involves a treatment group and a control group. (we can also compare 2 treatments, with or without a control group)

Control

… do not receive the treatment

units

The individuals in experiments (also called subjects when they are people)

Observational study

The process of observing and measuring specific characteristics without attempting to modify the individuals being studied. Tells what happens and can’t describe cause-effect relationships. Accessing reliable records counts as this even if someone else previously did the actual observation

same individuals/units

When we are interested in studying the relationship between 2 variables, we measure both variables on the ….

response variable

The … measures an outcome of a study

explanatory variable

An …explains or influences changes in the response variable

influence

Observe individuals and measure variable of interest but do not attempt to … the response

passive

In observation studies, the investigator’s role is basically … Individuals in a sample are studied but no attempt is made to manipulate or influence the responses

2 variables

Observational studies is good for establishing whether …. are related, or to learn characteristics of a population

response

Experiments deliberately impose some treatment on individuals in order to observe their …

active

In experiments, the investigator’s role is …Variables are manipulated, the study environment is regulated

experimental units

Treatments are applied to the …., to try to determine the effects of the treatment on the response variable

causation

The experimental study is better for establishing … (cause and effect)

Experiment, design of experiment, response variable, experimental unit/subject, treatment

What are the key terms used in experimental design?

effect

In designing an experiment, we wan to investigate the … of different treatments on a response variable

experiment

Process of collecting sample data

Design of experiment

Plan for collecting sample data

Response variable

Variable measured in experiment (outcome)

Experimental unit/subject

Object upon which the response is measured (or) individuals on which the experiment is done

Treatment

A specific experimental condition applied to the units or subjects

Treatment effects and experimental error. Lurking variables and confounding variables

There will be variability of responses for 2 basic reasons…

Treatment effects

This is what we’re looking for, different treatments causing different outcomes

Experimental error

1. Variability among observed values of the response variable for experimental units that receive the same treatment

2. We want this to be as small as possible

Lurking variables

A variable that is not among the explanatory variables in a study and yet may influence the interpretation of the relationship among response and explanatory variables

Confounding variables

Two variables are confounded when the effects on the response variable cannot be distinguished from each other

Control, randomization and replication

What are the principles of experimental design?

Control

Control the effects of lurking/confounding variables and other sources of variability on the response by carefully planning the study. Receives no treatment and is used a baseline or or comparison for the treatment group

Randomization

Randomly assign experimental units to treatments to reduce or eliminate bias. The use of chance to divide experimental units into groups. Refers to the methods used to assign those already in a sample to a treatment group

Replication

Measure the effect of each treatment on many units to reduce chance variation in the results

Sampling methods

Sampling refer to the methods used to collect people/units to be in a sample

Completely randomized designs, randomized block designs, matched pairs designs

What are some randomization techniques?

Completely randomized designs

Participants are randomly assigned to treatments (including control groups). By randomly assigning subjects to treatments, the experimenter assumes that, on average, lurking variables will affect each treatment group equally; any significant differences between groups can fairly be attributed to the explanatory variable.

Randomized block designs

The experimenter divides participants into subgroups called blocks, such that the variability within blocks is less than the variability between blocks. Then, participants within each block are randomly assigned to treatment groups. (Blocks may be age groups, sex, location, etc.)

Matched pairs designs

A special case of the randomized block design. It is used when the experiment has only two treatment groups; and participants can be grouped into pairs, based on one or more blocking variables. Then, within each pair, participants are randomly assigned to different treatments. This can also be done as a before-and-after experiment where the same subject is recorded before and after the treatment.

Placebo

A false drug or treatment that the subjects believe is real (examples include sugar pills, saline solutions, fake treatment, etc.

Placebo effect

The tendency to react to a drug or treatment regardless of its actual physical function

skewed

When only the treatment group is given a drug or other treatment, the results may be … by the placebo effect. Some individuals may respond to the treatment and others may only be responding because they believe it will work.

both

If both groups receive a treatment, whether real or fake, then … groups may experience the placebo effect

Bias of the subject and bias of the researcher

What are some issues in designing an experiment?

Bias of the subject

Similar to response bias in sampling, subjects may want to please the researcher or hope for a specific outcome.

Hawthorne effect

When people behave differently because they know they are being watched

Bias of the researcher

People subconsciously behave in ways that favor what they believe. Researchers, even when following a protocol, are no different. They may assign subjects to groups or report results in a biased way. They may treat people or animals differently when holding certain expectations of their treatment.

Blinding

Blinding is when individuals associated with an experiment (as a subject or experimenter) are not aware of how subjects have been assigned (treatment or control, treatment or placebo). Without this knowledge, the subjects are less likely to respond with bias and the researchers are less likely to allow their biases to influence the study.

Single-blind study

Those who could influence the results (subjects, administrators, technicians, etc.) are blinding

Double-blind study

Those who evaluate the results (judges, physicians, analysts, etc) are blinded as well

necessary or reasonable

Blinding should be used if possible, but in some situation is not….

No

Can we (reasonably) blind the subjects?