Untitled Flashcards Set

1. Why is there another test statistic besides z?
When should t be used instead of z?
Using z for hypothesis tests is appropriate (and a very good idea!) only
when the population standard deviation (σ) is known. If this is not
known, t should be used instead.
2. What kind of design is the single-sample t-test
appropriate for?
When the research design has a single sample of subjects who each
provide one score, and the research question is whether the population
from which the sample was drawn is the same as or different from some
known (or theorized) population value, a single-sample t-test should be
performed.
3. Why are critical values for t different than for
z?
Unlike z, which is based on one distribution (the normal curve), t is
based on many distributions which are similar to but not quite the same
as the normal curve. The more dfs there are in a sample, the more the t
distribution is like the normal curve.
4. What is df? Degrees of freedom. For a single-sample design, it equals n – 1. As for
what it means conceptually, it's a value that represents how many scores
are "free to vary" in a sample (see pp. 95-96).
5. What is Cohen's d? When and why should it be
computed?
Cohen's d is a measure of effect size, and gives us an idea of how much
of an effect a treatment has. It should be used as a supplement to a
hypothesis test when the null hypothesis has been rejected.
6. What assumptions should be true to carry out a
hypothesis test for a single-sample design with
t?
1. The subjects in the study have been randomly sampled from a
population.
2. All scores have to be independent of one another.
3. The population from which the sample was drawn must be
normally distributed (or n must be greater than about 30).
7. What is estimated standard error (sM)? Standard error tells us the typical distance between a population mean
(μ) and the mean of a sample (M) drawn from that population.
Ch 10
1. What kind of design is the independent-
samples t-test appropriate for?
When the research design has two different samples of subjects, who
each provide one score, and the research question is whether the two
populations from which the samples were drawn are the same as or
different from one another, an independent-samples t-test should be
performed.
2. What is pooled variance (s2p)? Why compute
it?
Pooled variance is a measure that combines variance from two samples
together. We compute it because two estimates of variance are better
than one – that is, pooled variance allows us to take advantage of having
two samples instead of just one.
3. What assumptions should be true to carry out a
hypothesis test for an independent-samples
design with t?
1. The subjects in the study have been randomly sampled from
one of the two populations in question.
2. All scores have to be independent of one another.
3. The two populations from which the samples were drawn must
be normally distributed (or n1 + n2 must be greater than about
40).
4. The two populations from which the samples were drawn must
have equal variances (this is what justifies pooling variance).
4. What is estimated standard error (s(M1 – M2))? One way to think about is that it's the (estimated) typical distance
between (M1 – M2) and (μ1 – μ2). This is kind of abstract, though. If we
assume that the null hypothesis is true (that is, μ1 – μ2 = 0), it becomes a
little easier to think about – it's the typical distance between two sample
means taken from the same population. Think about it this way: If you
have a population with a mean of 50, you might take one sample and
find its mean, and then take another sample and find its mean – they
probably won't be the same, and the standard error tells you how far
apart they would typically be.

Ch 11
1. What kind of design is the related-samples t-
test appropriate for?
When the research design has one sample of subjects, but who each
provide two scores, and the research question is whether the two
populations that the scores represent are the same as or different from
one another, a repeated-measures (related-samples) t-test should be
performed.
2. What assumptions should be true to carry out a
hypothesis test for a related-samples design
with t?
1. The subjects in the study have been randomly sampled from a
population.
2. All difference (D) scores have to be independent of one
another.
3. The population from which the sample was drawn must be
normally distributed (or n must be greater than about 30).
3. What is estimated standard error (sMD)? Standard error tells us the typical distance between a population mean
(μD) and the mean of a sample (MD) drawn from that population.
4. Why is a related-samples t-test more powerful
than an independent-samples t-test?
Because finding difference scores removes most of the individual
differences that exist between subjects. This in turn leads to a lower
standard error, which leads to a larger tobs.
Confidence intervals
1. What is estimation? Estimation is an inferential statistical technique used to try to come up
with an estimate of a population parameter based on a sample statistic.
We are usually concerned with estimating a population mean (μ) based
on a sample mean (M).
2. What is an interval estimate? An interval estimate is range of values that forms an interval in which
the population parameter is thought to lie (although the confidence we
have in this varies).
3. How is an interval estimate made? The following formula can be used to make an interval estimate for a
single sample using t (there's a similar formula for z):Mcrit stM 
M is, as usual, the mean of a sample. t in this case is the t-score
associated with the middle 95% (or 90% or 80% or whatever level of
confidence is desired of the interval) of the t distribution.
4. What is the most-typical level of confidence
associated with an interval estimate?
95% is the conventionally preferred level of confidence.
5. What does "95% confidence" mean? This is complicated. The idea is this: If a sample is taken from a
population over and over and over again, and for each sample a 95%
confidence interval is computed, 95% of the time the confidence interval
should contain the actual mean of the population. It does not mean that
there is a 95% probability that the population mean is in the interval; it's
either in the interval or it's not.

Ch 12
Why do an ANOVA? To find out if there are significant differences between two or more
population/treatment means.
What does "significant" mean when an ANOVA is
performed?
"Significant" means that the differences between means (that is, between
conditions in an experiment) are more than would be expected by
chance alone. The idea is that even if there are no differences between
conditions, because of sampling error we would expect sample means to
be at least a little difference. Significance testing allows us to see if
means are different enough to be convinced that the differences are real.
What do the results of an ANOVA tell us? That is, what
does rejecting the null tell us?
That there are or are not differences somewhere between
population/treatment means.
What further analyses need to be done if the null
hypothesis is rejected?
Posttests, like Tukey's HSD.
Why do posttests need to be done? To find out which means are different (and draw more-specific
conclusions). The ANOVA alone does not tell us this.
Should I do posttests if the null hypothesis has not been
rejected?
No. Not rejecting the null indicated that you are essentially concluding
that there are no difference among the means that are worth discussing.
How do you measure effect-size when you perform an
ANOVA?
η2 = SSbetween / SStotal
What does η2 mean? It tells how much of the variability in the dependent variable is
accounted for by differences between treatments. It is a proportion and
therefore has to be between 0 and 1.
Are there guidelines for deciding if η2 is small, medium,
or large?
Yes! .01, .09, and .25 are conventional values for labeling η2 small,
medium, and large, respectively.
What do you mean by "partitioning" variability or df? SStotal is chopped up into SSbetween and SSwithin. The same is true of dftotal.
What does MS mean? MS means mean square (or "mean of squared deviations"), but it's really
just another name for variance.
What does MSbetween measure? MSbetween measures variance between groups.
What does MSwithin measure? MSwithin measures variance within groups.
Why is the F-ratio called a ratio? Because it's the ratio of two MSs, or variances.
What makes MS or SSbetween go up or down? As group means get farther apart, SSbetween goes up. As they get closer
together, SSbetween goes down.
What makes MS or SSwithin go up or down? As the scores within groups get more spread out, MS and SSwithin go up.
As the scores in groups get less spread out, MS and SSwithin go down.