Masterclass 1-3 Flashcards
What is a variable?
A variable is a characteristic of things that may take more than one value. In probability, a random variable is an expression whose value is the outcome of an experiment.
What are the two components of Hume’s conception of causality?
- CONTIGUITY (contiguity in space and time of cause and effect) and
- CAUSE-EFFECT (cause occurs before effect), where there is a necessary connection if cause ad(?) effect.
What is a sufficient condition for an effect to occur?
if that condition occurs the effect will occur. It may be necessary or not.
What is a necessary condition for an effect to occur?
If and only if that condition occurs the effect will occur. It may be sufficient or not.
What are necessary and sufficient conditions?
With both necessary and sufficient conditions, they have to occur for the effect to occur and no other conditions will make the effect occur.
Why did David Lewis propose to take into account only the second part of Hume’s definition of causality?
There are situations in which the first rule occurs and there is no causality: night follows the day but the day is not the cause of night. The connection between cause and effect may not happen repeatedly (e.g. a person shoots and kills another person).
What does each rug of Pearl’s ladder of causation tell us about causality?
Observations: correlations;
Actions: causation by intervention;
Imagination: causation by modelling
What is a chain? And, why is it important?
In chain and mediation of X–>V–>Y,
The effect of X on Y is mediated by V.
What is a common cause? And why is it important?
In confounding U–>X–>Y, we are studying whether X causes Y. But U–>Y and so U is a common cause of X and Y.
U is confounding the link X–>Y (the confound is on the DV Y).
What is a collider? And, why is it important?
In Selection bias X–>Y–>Z, we study whether X causes Y (Z the collider is on BOTH the IV X & DV Y).
Z is a common effect of X and Y or a collider. If we control for Z we may find an association X–>Y that was absent if not controlling. Or we dilute an actual r/s.
In a study, the Bayes Factor (BF10) comparing an H1 to a H0 is 0.05…
Meaning: the data is 20x more likely under the H0 than under the H1.
The Bayes Factor (BF10) comparing the H1 and H0 is 0.10…
Meaning: the data is more likely under the H0 than under the H1.
The prior distribution in the model of the H0 used by JASP is:
A spike over the value 0.
In a study, the Bayes Factor (BF10) comparing an H1 to a H0 is 0.5…
Meaning: the data is 2x more likely under the H0 than under the H1.
The probability of the data under the H1 is 0.001 and that of the H0 is 0.0001. What is the value of the Bayes Factor (BF10)?
a. 1
b. 0.0001
c. 10.
d. 0.001.
The prior distribution in the model of the H1 used by JASP is:
A symmetrical distribution centered around the value 0 (or another H0).
In the H0 testing approach, if the value of the statistic calculated in the sample (e.g. b1), or a more extreme value, has a lower probability than that of a pre-established threshold, what is the correct decision?
To reject the H0.
In the Maximum likelihood estimation approach, why is the more complex model penalised?
To avoid overfitting.
Which component in the Bayesian approach is the most similar to an equivalent component in the Traditional approach?
Probability of the data given a parameter value.
The probability of the data under the H1 is 0.008 and that of the H0 is 0.8. What is the value of the Bayes Factor (BF01)?
a. 0.0001.
b. 100.
c. 1.
d. 0.001.
Consider two sampling distributions (A and B) from the sample population of values: A was constructed with samples of 10 values, B was constructed with samples of 100 values.
Ans: The standard error of A is larger than the standard error of B.
The sampling distribution of the mean is constructed by generating…
Thousands of samples of values and calculating their means.
Prior knowledge combined with likelihood allows to calculate…
A posterior distribution.
The mathematical formula that corresponds to the Linear model with one predictor variable:
Y ~ Normal (Beta0 + Beta1 X, sigma)
The standard deviation of a population and the standard error of the mean calculated from that population tells us that:
The standard deviation of the population is higher than the standard error of the mean.
If the prior distribution for a parameter value is a normal distribution…
the parameter values close to the mean of that distribution are more plausible than those far from the mean of that distribution.
The mathematical formula that corresponds to the Simplest linear model?
Y ~ Normal (Beta0, sigma)
If the prior distribution for a parameter value is a uniform continuous distribution…
All the possible values of that parameter are equally plausible.