You can do it, believe in yourself
v Cover page - 800547-B-6/Statistics for PM CIS (Fall)
v Name of the course: Statistics for PM CIS (Fall)
v Course code: 800547-B-6
v Snr. 2142596, Anr. 509637
v Name: Yen-Ju, Lin
v The exam consists of 5 questions contributing to a total of 68 points. This you also already know.
v The 5 questions are made up out of a couple of sub-questions each (e.g., Q1 consists of Q1a-Q1e).
This means that, in total, you will have to give 18 separate answers. Some of these are very short
(just providing 1 number), some of these are a bit longer but still short (a couple of sentences),
some are quite long (a report). On average, you have 180 mins/18 answers = 10 mins/answer.
Remember that these 10 minutes CONTAIN EVERYTHING, including reading the questions and
case descriptions and conducting the analyses.
v The (sub)questions are a mix of exercises and theory questions, very similar to what you got in
the midterm. In particular:
v Question 1 provides a dataset and case description. It requires you to conduct some analyses and
answer 5 subquestions, all with short answers (a few sentences max.).
v Question 2 requires you to write an APA report on the same dataset.
v Question 3 also requires you to write an APA report on the same dataset.
v Question 4 provides a new dataset and case description, after which you have to answer 6
subquestions with short to very short answers (1 number, a couple of numbers, or few sentences
max.).
v Question 5 is theory. You will get 5 statements. For each, you will have to decide whether you
agree/disagree (or whether it makes sense/does not make sense; if it's true / false,...) and briefly
explain why (a few sentences max.)
,GUIDELINES FOR REPORTS: READ THIS CAREFULLY
1. If any variable is based on several items, describe what is measured, on
which scale the items were measured, and whether or not the scale is
reliable;
2. In case you’re asked to write a full APA report, always start the report with
(re-)phrasing the hypothesis that you are going to test and name the test
that you use;
3. For all reports, be as exhaustive as possible, interpreting all relevant
statistics and assumption checks for the case at hand;
4. Unless indicated otherwise, always test and report on assumptions of your
analysis (even if the assumption has been met). Hence, for this exam, also
report the z-scores of Skewness and Kurtosis when the assumption of
normality has been met (Note: if > ± 1.96, then p < .05);
5. If the assumptions have been violated, report on that!
6. If the sample size is above 40 you MUST rely on the central limit
theorem to address violations of the assumption of normality (but still
report the assumption tests and your conclusions!)
7. Always give basic descriptive statistics (e.g. M, SD, Median if applicable) for
all relevant variables;
8. Follow APA standards to report numbers:
a. Always round numbers to 2 decimal places, unless it's a value smaller
than .01 (e.g., P = .001).
b. Only report leading zeroes (0 at the start of a number) if values smaller
than 0 or higher than 1 are theoretically possible (e.g., P = 0.001 is wrong,
but Mean = 0.005 is ok).
9. Report the direction and the magnitude of the relationship/effect (effect
size);
10. Draw a conclusion about whether the results support the hypothesis or
not;
11. Always use a two-tailed test unless the question specifically states
otherwise.
Wilcoxon test: is a non-parametric statistical test used to compare two related samples or matched
pairs to assess whether their population mean ranks differ. It can be applied in situations where the
data do not meet the assumptions of a parametric test (e.g., normal distribution)
Shapiro-Wilk: Determine whether a dataset follows a normal distribution. It evaluates the null
hypothesis that the data is normally distributed. A significant p-value (typically < 0.05) suggests that
the data deviates from normality
Mann-Whitney U test: Non-parametric test used to compare differences between two independent
groups when the assumptions of parametric tests (like the t-test) are not met, particularly the
assumption of normality.
Omnibus test: comparing all means together
,Post-hoc tests and contrasts: comparing 2 particular means
Generalized Linear Model
v T-tests, ANOVA, Correlation and Linear Regression belong to GLM
, Midterm Questions
v Which of the three measures for the central tendency represents the distribution of ‘mean_fake’
best?
t Answer model
l 1pts for choice of median
l 1pts for explanation: Right-skewed distribution; mean is affected by the left skew.
Median provides a better representation of the center of skewed data, because it is
just the midpoint in a rank of set distribution.
v Standard Error is the standard deviation of (the sampling distribution) of the mean.
v CLT guarantees that the sampling distribution of statistics such as the sample mean will be
normally distributed, not the variables themselves.
v Type I and Type II errors
t It’s not that raising n raises the Type I error probability. Multiple testing and / or changing the
significance level would do so. Also, the higher the n is, the lower the probability of Type II
error is. Type I error is not dependent on the sample size.
------------------------------------------------------------------------------------------------
Questions from Practice exams
v From the information given in the abstract, formulate a hypothesis that can be tested using an
independent-samples t-test. Note: the independent and dependent variables need to be mentioned
in the abstract as they are (2 pts)
t The communication effectiveness rated by participants who were presented health messages
from a {different source of information (medical professional, or a layperson)} is different.
v State the null hypothesis that would be used in a null hypothesis significance test for the
hypothesis you just formulated (in 1a). (1 pt)
t There is no difference between {the communication effectiveness}rated by{participants who
were presented health messages from a medical professional} and{those who were presented
health messages from a layperson}
v What is the dependent variable in your hypothesis? What is its measurement level? (1.5 pt)
t The dependent variable is the communication effectiveness on a 5-point, 10-item scale. The
measurement level is {continuous, interval.}
v What is the independent variable in your hypothesis? What is its measurement level? (1.5 pt)
t The independent variable is{the source of the information}with two levels:{ expert and non-
expert.}The measurement level is {categorical, binary.}
v Suppose you want to compare two groups of participants on a certain attitude score. You find that
the distributions in both groups look bell-shaped but have several scores that form a tail toward
the lower end.
(1) How do we call this pattern in statistical terms?
(2) Assuming that the total sample size is only n = 25, what would be the best way to deal with
this pattern in your analyses?
(3) Had the sample size been n = 500, would there have been another way to deal with the issue?
(2.5 points)
1) Skewness (0.5 pt)
2) Run a non-parametric test that does not require normality: Mann-Whitney U. (1pt)
3) In this case, we could have relied on CLT to argue that the errors (and the sampling distribution
of the mean difference) would still be normally distributed regardless of the distribution of the
scores in the groups. (1pt)
v Mr. Worry performs a regression analysis and finds that the intercept of his regression equation is
-12. The regression analysis predicts the weight of a person, which cannot have a value of -12.
Now Mr. Worry thinks that something must have gone wrong. Reassure Mr.Worry that
everything is fine. (3 points)
