adv 309r exam
types of experiment's external validity
- population-related external validity - setting-related - measurement-related
pretest to post-test with control
- two groups with four measures - makes certain there is equivalency between the treatment and control groups before the start of the research - An alternative to the post-test only with control design - Both designs control for premeasurement, history, maturation, and instrumentation threats - Controls selection threats - Weakness: doesn't eliminate mortality threats - *true experimental*
pearson's correlation coefficient range
-1, 1
t-test process
1. check assumptions 2. calculate t-statistics 3. conclusions
requirements for causality
1. events must take place in the proper order 2. events must take place at the same time and show an explicit relationship 3. alternative explanations must be reduced and eliminated whenever possible 4. strength of association (p value)
characteristics of experiments
1. identify what you need to learn 2. take the relevant actions (conduct the experiment by manipulating one or more variables) 3. observe the effects and consequences of those actions on other variables, and then 4. determine the extent to which the observed effects can be attributed to actions taken
anova assumptions
1. samples are random and independent of each other 2. the populations are normally distributed 3. the populations all have the same variance
two variable design (i.e., factorial design)
1. study two or more factors 2. crossed factors are arranged in a factorial design
a researcher wants to compare media use among four different ethnic groups. What statistical test should be applied?
ANOVA
chi square
an examination of a pattern of responses to see the difference between the expected and the observed result - when data represents frequency or percentages (nominal level data) - subjects in each category are independent - when there is only one independent variable
mortality threats
arise when respondents drop out of a study btwn the pretest and the post-test and as a result, it cannot be determined whether differences between the two tests are the results of the experimental treatment or are the result of different group characteristics at the two stages of measurement
t-test
assignment to one of no more than two groups (independent variable)
sources of variability in an anova
between group and within group
solomon four-group design
both the most powerful and the most-resource intensive experimental design - Controls for all major threats to an experiment's internal validity - Combines a pretest to post-test with control design with the post-test only with control design - Weakness: complex, costly, takes longer - *true experimental*
f-ratio
btwn group variability/within group variability
experiments
can be used to bypass recall and actually observe behaviors
testing threats
can result from repeated administrations of the same questionnaire or survey, when it is administered once before the treatment, and once again after the treatment - a consumer may know how to answer the test questions "correctly" on the second try
in a five group comparison study, greater within group variance will __________, your f-ratio by definition
decrease
internal validity
determines the extent to which we have confidence that observed results are due to the experimental manipulations
anova is restricted to 3 or 4 group comparisons
false
r=1 is stronger than r=-1
false
f-ratio score tells you what means are different
false: it tells us that they are different not which are different
R-squared
gives you the variance explained in the dependent variable - always expressed as a percentage
two group post-test with control
has two groups, one with pretest and posttest and one just with a post-test - The main problem with this is that there was no true random assignment so there is no way to tell or even assume any equivalence in pretest scores - *quasi*
true experimental designs
have a control group and use random assignment to form test and control groups 5 most common types: 1. simulated pretest to post-test 2. post-test only with control 3. pretest to post-test with control 4. solomon four-group design 5. factorial designs
in a factorial design, the two factors are considered the
independent variables
external validity
influences the extent to which the experimental results are generalizable to the "real world" - population related - measurement related - setting-related external validity
quasi-experimental design
not true experiments; fail to eliminate a large number of threats to internal validity common types: 1. one group post test only 2. one group pretest to post-test 3. two group post-test with control
factorial design
observes the effects of two or more independent variables at the same time where each of these variables has two or more levels or aspects - Most appropriate if you want to see the effect of two independent variables working together
interaction threats
occur whenever an interview administered before the start of the experiment affects a respondent's sensitivity or responsiveness to the independent variable - occur whenever the independent variable is more likely to be noticed and reacted to then it would be without exposure to the initial measurement instrument
selection threats
occur whenever, at the start of the experiment, the tests and control groups differ in terms of relevant demographics, attitudes, or behaviors
researcher bias threats
occurs whenever the actions of the experimenter affects the outcome of an experiment
a company wants to compare their latest commercial against a pool or prior commercials. what statistical test should be run?
one sample t-test
inferential statistics
provides the basis for objectively determining the amount of confidence you can have in drawing conclusions about differences and relationships observed in a set of data
history threats
refers to any events or influences beyond those intentionally manipulated by the researcher that occur during the experiment and that have the potential to affect the experimental outcome as measured by the dependent variables - ex. weather
instrumentation threats
refers to changes made to the measurement instrument or data recording techniques during the experiment
surveys
respondents are asked to recall and report on their behaviors
maturation
respondents attitudes, behaviors, and physiology change during an experiment - all of these have the potential to affect and distort the levels of the dependent variable
one group pretest to post-test
similar to the one group post-test only except there's a pretest - assessed by comparing the levels of the dependent measure in the post-treatment to the levels in the pretreatment measures - no control for any historical threats to internal validity - premeasurement threats -*quasi*
one group post-test only
takes a group of individuals, exposes them to the treatment or experimental manipulation (the independent variable), and then measures the dependent variable(s) as part of the post test - you have to use judgment since there is no control or reference group - fails to control for several additional threats to internal validity, specifically maturation, selection, and mortality - *quasi*
two variable main effect
the change in response produced by a change in the level of the factor
interaction
the combinations of main effects on the dependent variables
main effects
the separate influence of each independent variable on the dependent variable
goal of an experiment
to determine causality- the effect of changes in one area on one or more other areas. This is important because its more accurate than descriptive research in making decisions on advertising
"R" assumes that one variable predicts the other
true
"r" simply assumes the two continuous variables are related
true
correlation and simple regression are about looking at relationships between two variables
true
multiple regression is appropriate when there is more than one independent variable and one dependent variable
true
pearson's correlation coefficient: "r" shows linear correlation between two continuous variables
true
hi and low media use is though to affect attitude toward advertising. to test this what statistical test should be applied?
two sample t-test
paired t-test
used to compare the means of two dependent samples
two-sample t-test
used to determine whether the mean of one group is equal to, larger than or smaller than the mean of another group
one-way anova
used to determine whether three or more populations have different distributions
one sample t-test
used to test whether the population mean is different from a specified value
simulated pretest-post-test
uses one group of randomly assigned respondents for the pretest measurement and a second group of randomly assigned respondents for the treatment and post-test measurement - controls for pre measurement and interaction threats - The simulated pretest and post-test design estimates treatment effects by comparing one group's pretest to a different groups post test - Weaknesses: doesn't control for any of the other threats like history, maturation, instrumentation, and selection - *true experimental*
post-test only with control
utilizes two groups of respondents and two measures; compares exclusively post-test measures, one measure obtained from the treatment group and one from the control - Controls for a greater number of threats to internal validity - Controls for history, maturation, and instrumentation, as well as premeasurement, and interaction threats - Weakness: doesn't eliminate mortality threats - *true experimental*
equal variance
variances are approximately the same across the samples - you have two different means: if the variation is relatively equal, that is one sample variance is no larger twice the size of the size of the other, then you can assume this
independent variable
what the experiment manipulates
dependent variable
what the researcher is interested in explaining, and as a result, is the measure used to evaluate the influence of the independent variable
premeasurement threats
whenever an interview administered before the start of an experiment has a direct effect on a respondent's attitudes, actions, or behaviors during the experiment - occurs without exposure to the independent variable - all observed attitudinal or behavioral changes are the result of exposure to the initial measurement instrument
difference between r and R
with r, the two variables are treated as equals. in regression (R), one variable is considered the independent (predictor) variable and the other the dependent (outcome) variable