Transcript Section 1

The Research Skills exam:
The four horsemen of the apocalypse: pestilence, war, famine and the RS exam.
The exam format:
Open book - you can use your notes, books and
an approved calculator in the exam.
Two hours and five sections Section 1: basic concepts.
Section 2: pick a test and use it.
Section 3: which test?
Section 4: pick a test and conclusions
Section 5: write a results section.
Preparation is everything:
Open book - make sure your notes
are useable under conditions of
stress!
Are they clear?
Do you know where everything is?
Write clear, step-by-step guides to
calculations.
Be strategic – don’t start at page 1
and work through. Answer all the
easy questions first, then tackle
tricky stuff if there is time.
Section 1: Basic concepts
Ten multiple choice questions, testing knowledge
of basic statistical concepts.
Revise by writing clear definitions of terms such
as "mean", "standard deviation", "normal
distribution", "ratio data", "one-tailed test", etc.
Read all of the alternatives carefully, before
making your choice.
1
The Standard deviation is:
(a) a statistic that tells us how well our sample mean
is likely to reflect the true, population, mean.
(b) a statistic that tells us how scores are distributed
around the mean of the set of scores.
(c) a statistic that tells us whether scores are
normally distributed.
(d) a statistic that tells us what is the most
commonly-occurring value in a set of scores.
(e) none of the above.
Correct definition [ b ]
2.
p < .05 means:
(a) the obtained result is not due to chance.
(b) the obtained result is a fairly important effect.
(c) the obtained result is likely to occur by chance
95% of the time.
(d) the obtained result is likely to occur by chance
less than 5 times in a hundred.
(e) the obtained result is likely to occur by chance 5
times in a hundred.
Correct definition
[d]
Section 2: Pick a test and use it:
Given a set of data, you have to
(a) decide which test is appropriate to use;
(b) perform your chosen test on the data.
Only one of the following tests will be correct:
Independent-means t-test
Repeated-measures t-test
Wilcoxon
Mann-Whitney
Friedman's
Kruskal-Wallis
Spearman's rho
Pearson's r
Chi-Square
Are there sex differences in car
parking ability?
9 women and 9 men park their
cars while the experimenter
times how fast they do it.
subject
time to park: (sec)
subject
time to park (sec)
male 1
30
female 1
66
male 2
45
female 2
65
male 3
48
female 3
67
male 4
55
female 4
73
male 5
34
female 5
58
male 6
28
female 6
41
male 7
44
female 7
59
male 8
37
female 8
62
male 9
52
female 9
54
(a) Pick the right test:
Scores or frequencies? Each participant provides a score.
Eliminates Chi-square
Relationships or differences? Men versus women.
Eliminates Pearson's r and Spearman's rho
Two, or more than two conditions? Men vs. women.
Eliminates Kruskal-Wallis and Friedman's
(a) Pick the right test:
Independent or repeated measures? Men vs. women.
NOT Wilcoxon
M/W
Independent t
NOT Matched t
Ordinal, interval or ratio data? Time to park is ratio
NOT Wilcoxon
NOT M/W
Independent t
Matched t
Data normally distributed? No reason to think not!
NOT Wilcoxon
NOT M/W
Independent t
Matched t
Homogeneity of variance? No reason to think not!
NOT Wilcoxon
NOT M/W
Independent t
Matched t
Correct test: Independent-means t-test.
(b) Perform this test on the data,
Report the test statistics clearly and in the proper
format, and say what the results mean in words.
"An independent-means t-test was performed on
these data. This revealed that women and men
differed significantly in the amount of time they
took to park their cars (men: mean = 41.44 sec,
s.d. = 9.65 sec; women: mean = 60.56 sec, s.d. =
9.23 sec: (t(16) = 4.29, p < .001)".
Section 3: Which test?
On the basis of a brief description of an
experiment and some data, you choose the
most appropriate test. For each question give
the letter corresponding to the correct test.
A. Wilcoxon
G. One-way independent-measures ANOVA.
B. Friedman
H. Spearman's rho
C. Mann-Whitney
I. Pearson's r
D. Kruskal-Wallis
J. Chi-Squared
E. Unmatched t-test
K. One-way repeated measures ANOVA
F. Independent means t-test
1. The statistics exam scores of three groups of
students were recorded: group A had revised for
5 hours, group B for 10 hours and group C for 20
hours. What test is required to test the hypothesis
that revision time affects exam performance?
Test: [G (one-way independent-measures ANOVA]
2. The number of students passing or failing a
statistics exam in each of three groups of
students was recorded: group A had revised for 5
hours, group B for 10 hours and group C for 20
hours. What test is required to test the hypothesis
that revision time affects exam performance?
Test: [J (Chi-square]
Section 4: Pick a test and conclusions
Here are the instructions for this section:
In the following questions you are given some details of an
experiment, the results of a number of statistical tests, and a
set of conclusions. Only ONE of these tests is appropriate, and
only ONE in each set of conclusions is correct. Thus only
TWO of the statements are correct in each of the following
questions. Indicate which two are correct by writing the
appropriate letters in the table at the end of this section.
Look at the data; select the most appropriate test
statistics; pick the correct conclusion for those results.
A researcher is interested in whether a drug affects appetite in rats. Eight rats were
tested twice, once with the drug and once without the drug, in a random order. In
order to feed, the rats need to climb a slope to get a food pellet, and the number of
times each animal climbed the slope in a 5-minute period was recorded.
Statistical tests:
(a) Matched sample t-test: t (7) = 5.11, p = .001. (CORRECT)
(b) Independent measures t- test: t (14) = 2.60, p = .02.
(c) Mann Whitney U (8,8) = 11.00, p = .03.
(d) Pearson’s r = .77, p = .02
(e) Spearman's rho = .80, p = .02
(f) Wilcoxon test: z = 2.52, p = .01.
Drug:
No drug:
10
15
15
22
11
18
22
24
18
19
Conclusions:
14
22
(g) The drug makes rats’ performance more variable.
16
19
(h) Rats climb the slope significantly faster after taking the drug.
(i) Rats climb the slope significantly more often after taking the
19
23
drug.
M = 15.62
M = 20.25
(j) Rats climb the slope significantly less often after taking the
drug. (CORRECT)
SD = 4.03*
SD = 3.01*
(k) There is a significant correlation between the number of times
that rats climb the slope with and without the drug.
(l) The drug makes the food pellets significantly more tasty to the
rats.
*using n-1 SD formula
Section 5: Write a results
section
Revision
Revision + gin
5
9
Read the scenario supplied; write a
results section; interpret the results
appropriately:
14
7
13
6
17
15
The hypothesis is that revision combined
with alcohol leads to better exam
performance than revision alone. This was
assessed by measuring subjects' exam
performance, after either 1) subjects had
revised for 5 hours a week while drinking 500
ml of gin; or 2) simply revising for 5 hours
each week. The results in terms of the exam
scores are shown in the table below, together
with the means, standard deviations, and t
statistic for the unmatched t -test used to
compare the scores.
12
5
11
6
19
12
11
13
M =12.75
M = 9.13
SD = 4.23*
SD = 3.76*
(NB: maximum exam score is 20).
t = 1.81
(*NB: using the n-1 version of the
SD formula)
(a) Graph the data in a form appropriate for inclusion in a lab report:
This doesn’t have to be a work of art, but should be clear and
labelled correctly. Include error bars and a title! Standard error =
SD divided by square root of the number of subjects. Here, SEs
are 1.40 (4.23 / square root of 8) for the sober revisers and 1.24
(3.76 / square root of 8) for the boozed ones.
mean exam score (+/- 1 s.e.)
Fig. 1: effects of alcohol and
revision on exam performance
16
14
12
10
8
6
4
2
0
without gin
with gin
revision condition
(b) Describe the data in a form appropriate for
inclusion in the results section of a lab report
(assuming no graph in the results section).
"An independent means t-test was performed
on these data. This revealed that there was no
significant difference between the mean exam
scores of subjects who combined revision with
alcohol consumption and those who revised
without it (t(14) = 1.81, p > .05. Revision only:
mean = 12.75, s.d. = 4.23 sec; revision plus
alcohol: mean = 9.12, s.d. = 3.76 sec). Exam
performance is not significantly enhanced by
combining revision with alcohol consumption."