Final Review Flashcards
Who proposed the scales of measurement?
S. Stevens in 1946, a Harvard psychologist
What are nominal scales?
how many …belong to each category (categorical)?
What are ordinal scales?
the sequence of categories has meaning (e.g. Grades)
What are interval scales?
an equal-interval but no true zero (e.g., temperature)
What are ratio scales?
an equal-interval with a true zero point
Why are scales of measurement important?
it determines the type of statistical analyses possible
What are the conditions of the scales of measurement?
does it conform to the abstract number system? does it include identity? order? equal distance? an absolute zero?
What is a sampling distribution?
a frequency distribution (equivalently, a probability distribution) of a sample statistic
What is the distribution of sample means usually referred to as?
the sampling distributions of means, or the sampling distribution of the mean
What type of distribution is a sampling distribution of means?
a probability distribution
What type of frequency of distribution of means is obtained from an unlimited series of sampling experienments?
a relative frequency distribution of means
What does a sampling distribution of means consist of?
a sample of size n randomly selected from the population
In his favour slide, what determines the population of IQ scores in 10 year-olds?
u = 100 o = 16 n = 64
Do we want the critical value of B to be high or low?
as low as possible to maximize the power and size of the alpha
power = ?
1 - B
What is power?
the odds of getting a significant result when the H0 is false
Do we want to maximize or minimize the distance between u0 and u1?
maximize to ensure a greater likelihood of significance/power
What is a preferred measurement of effect size in samples?
partial omega-squared because it is more accurate estimation of the population effect
When in doubt of effect sizes, choose…?
adjusted or partial (smaller df for higher power)
What is the standard normal distribution?
a mean of 0 and a SD of 1
How is a standard normal distribution normally designated as?
N(0,1) N = normal 0 = value of u 1 = value of o^2 N(u,o^2)
What percentage of population are within 2 SD of the mean?
95%
What population are within 1` SD of the mean?
68%
T-distribution is normally distributed and mean of zero?
yes
general t-scores are normally distributed with a mean of zero?
nah. generally not normally distributed
The discrepancy between the t-distribution and the z-distribution gets worse when…
“n” gets smaller (central limit theorem)
The two degrees of freedom impact the shape of the F distribution
n - 1, with a indicating area of significance
What is a calculation of the chi-square distribution?
x^2 = sum of (O - E)^2/E
What does a chi-square distribution with df 1 look like?
infinity and then lowers to almost 0 in graph (positive-skew)
What does a chi-square distribution with df 2 look like?
much lower than infinity and then lowers to almost 0 in graph (positive-skew)
What does a chi-square distribution with df 4 look like?
nears zero, goes up a little and then down for a positive skew near 0 at 11
What does a chi-square distribution with df = 8 look like?
nears zero, goes up a little until 7, and then down for a positive skew near 0 at 15
When o is know, what type of test?
1-sample z-test (confidence intervals)
When o is not known, what type of test if one sample?
one-sample t-test with CIs about the mean
When o is not known, what type of test if independent subjects?
2 independent samples t-test with CIs about the DIFFERENCE between means
When o is not known, what type of test if repeated subjects?
matched-pairs (2-dependent samples) t-test with CIs about the mean difference
When we want to explore whether the effects of different treatments on the DV measure, we use for 2 means
a t-test with 1 predictor/IV
When we want to explore whether the effects of different treatments on the DV measure, we use for 2+ means
an ANOVA with multiple IVs
Why use ANOVA instead of t-tests?
can look at several independent variables and does NOT inflate the TYPE I error rate
anova = SSTOTAL = SSTREATMENT + SSERROR, then
use a priori comparisons, otherwise posthoc tests
df for 1-way anova?
between group = k - 1
within group = N - k
total = N - 1
Using a priori comparisons, be sure that if a treatment isn’t used then
it isn’t used in further comparisons (diagonal of negative values for coefficients for treatments)
2-way between subjects design:
SStotal = SStreatment (SSa + SSb + SSab) + SSerror
2-way between subjects Summary table
dfa = (a-1) dfb = (b-1) dfab = (a-1)(b-1) dferror = ab(n-1) == note that degrees of freedom are only used for the subjects, not for the comparisons! dftotal = N - 1
The main effects should not be further examined when there is a significant interaction effect.
true
the presence of a significant effect limits the sense of the _____ effects
main
3-way ANOVA between-subjects summary table
dfa = (a-1) dfb = (b-1) dfc = (c-1) dfab = (a-1)(b-1) dfac = (a-1)(c-1) dfcb = (b-1)(c-1) dfabc = (a-1)(b-1)(c-1) dferror = abc(n-1) dftotal = N - 1
ANCOVA partitioning variance
SSTOTAL = SSTREATMENT + SSERROR (SSERROR + COVARIATE)
What rows in SPSS output do I use to calculate ANCOVA?
Source IV1 (=name)
Source IV2 (=name)
Error (error)
Total (corrected total****)
Repeated-measures ANOVA =
SStotal = SSbetween subjects + sswithin subjects (ssbetween treatment + sserror)
repeated-measures summary table
df(between subjects) = n - 1
df(treatment) = k - 1
df(error) = (n-1)(k-1)
df(total) = n*k-1
When doing a repeated-measures summary, what row has a F-value?
the treatment. nobody cares what the F is for between subjects…it is nearly always insignificant
What is the relation in correlation?
usually assumed to be a straight line, indicating a linear correlation or regression
What is a curved line in bivariate correlation?
a non-linear regression
how are correlations detected by the eye?
usually by a scatterplot
A way of predicting the value of one variable from another is
regression
Regression is a ________ model of the relationship between two variables
hypothetical
What type of model is a regression?
a linear one
the relationship of a regression is determined using
the equation of a straight line
What rae the two kinds of chi-square tests?
distribution shape and independence tests
What are the distribution shape tests?
goodness of fit/one-way classification test and homogeneity test
What are the independence tests?
contingency table tests (a*b tables)
What aren’t independence tests?
association tests…the latter would be more of a goodness of fit test or a “two-way classification” test
What do both types of chi-squre tests measure?
observed frequency of categories with an expected frequency of categories
what are expected frequencies usually derived from?
the null hypothesis, although not always. they can be derived from different theories too
What other condition should be met in spss summaries for dependent t-tests?
high correlation of paired-samples correlations
What other condition should be met in spss summaries for independent t-tests?
Levene’s test should be passed which assumed the equal variance assumed, and that null hypothesis is passed
What does the asterix to the right of the correlation indicates?
a statistically significant result at the 0.05 level, TWO-TAILED
to test a hypothesis regarding the distribution of observations in the population, choose values from what hypothesis?
the null hypothesis, for the parameters in the pop model to find out the sampling distribution of a sample statisc would look like if the null hypothesis is true
R-squared = ?
eta-squared
