Understanding Psychology as a Science Flashcards
Distinguish objective and subjective probability
Objective probability is the long run relative-frequency; the frequency of an event that you expect to get in the long run, subjective probabilities is the degree of conviction in a belief
What do objective probabilities apply to? (what they don’t apply to also)
They apply to collectives, not singular events (so not hypotheses)
If we symbolise data by D and a hypothesis by H, how is the probability of obtaining some data given a hypothesis written as?
P(D/H)
What is a common misconception regarding the hypothesis and P(D/H)
P(D/H) is not the same as P(H/D) there is not probability of the hypothesis because it is simply true or false.
What is the reasoning behind the Neyman and Pearson approach to hypothesis testing?
Statistics can’t tell us how much to believe a certain hypothesis. Thus, Neyman and Pearson set up decision rules for accepting or rejecting hypothesis in such a way that we will not often be wrong in the long-run.
What is meant by the rejection region?
values of t so extreme (or more extreme) that the probability of obtaining a t in that region is equal to α, if H0 is true. If our obtained t falls in this region, we reject H0.
What is meant by the alpha? (a)
the level of significance that is set in advance. It is the probability of obtaining a t value in the rejection region
What is meant by the beta? (b)
the proportion of times we accept H0 when it is in fact false: p(accepting H0|H0 false).
What is meant by power and how do we calculate it?
(1-b)
It is the sensitivity- the chances that we will find an effect given that H1 is true
Why do larger sample sizes have more power?
Larger sample sizes have more power because they are better approximations of the population.
What is specifity and how do we calculate it?
(1-a)
the probability of finding that there is no effect given that there is none: p(accept H0|H0)
What is meant by stopping rules?
conditions under which you will stop collecting data for a study.
Give a common rule for stopping rules
A common rule is to run as many participants as is traditional in that area. Some studies use the rule of collecting data until a significant result is found, this results in an alpha of 1.
When doing three tests at a significance level of 0.05, what should the P level be lower than for each test?
0.05/3= 0.017
Name 5 common misconceptions in NHST
- You have absolutely disproved H0 when p < α or absolutely proved it when p > α.
- You have found the probability of H0 being true.
- You can deduce the probability of HA being true.
- A 95% confidence interval has a 95% probability of containing the population value.
- You know the probability that you make the wrong decision if you decide to reject H0.
What does the Duhem-quine problem state?
it is not possible to test a single hypothesis in isolation as every hypothesis relies on several other hypotheses, theories and assumptions about the world.
What is meant by observations are theory-laden?
Expectations and assumptions about the world influence observations. Observations, in turn, influence hypotheses.
In this regard of observations being theory laden, what does a good theory contain?
A good theory makes these expectations and assumptions visible.
What is meant by the substantive hypothesis?
the hypothesis based on the previous research
What is wrong with research and statistical hypothesis?
They are often not aligned with the substantive hypothesis
What is required in order to connect scientific findings?
Theory
What is the effect of a precise theory on falsification?
More precise theories require fewer data to be falsified.
When are null hypotheses useful and how do they influence the hypothesis?
A null hypothesis is only useful for simple hypothesis testing and this influences hypotheses being formulated in such a way to fit null hypothesis testing.
What form of faulty argument texhnique is this? Explain
Straw-man; No difference between experimental designs is very unlikely given that the test has enough power. Therefore, accepting a theory on the basis of rejecting a null is not a stringent theory.
What is meant by Abduction?
Abduction is inference to the best explanation.
What are the three forms of circle reasoning?
- Repeating the premise.
- The premise presupposes the truth of the conclusion.
- The premise is logically or semantically equal to the conclusion.
What is meant by equivocation?
This is putting forward a conclusion using vague or ambiguous terms.
If the null hypothesis is true, then the p-values drift randomly, and can produce a significant result. how is this different to bayesian statistics?
In Bayesian statistics, the Bayes factor does not drift randomly but drifts towards the correct decision.
In classical studies what three influences are there on the conclusion which is not the case for bayesian statistics?
the stopping rules (1), the timing of explanations (posthoc test or not) (2) and multiple tests influence the conclusion.
Why can probability be assigned to a single hypothesis in bayesian statistics?
Bayesian statistics is a method of learning from prediction errors. It assumes that probability does not exist but only uncertainty, which has to be quantified in a principled manner.
Why can bayesian statistics investigate P(H/D) rather than P(D/H)?
The data drive an update from prior knowledge to posterior knowledge.
What can the bayes factor be seen as?
The Bayes factor can also be seen as the predictive
updating factor for the posterior belief. It is the ratio of likelihood where the likelihood refers to the probability of obtaining the data given the hypothesis
Give bayes rule as an equation
P(H\D)=P(D\H)*P(H)\P(D)
Does a high predictive updating factor in favour of the alternate hypothesis mean that the alternative hypothesis is better? Explain
The prior distribution determines the posterior distribution, therefore, a high predictive updating factor in favour of the alternative hypothesis does not necessarily mean that the alternative hypothesis is better. It only predicts the dataset X times better than the null hypothesis in this case.
Give the strength of evidence which corresponds to each range of bayes scores
1-3; Anecdotal 3-10; Moderate 10-30; Strong 30-100; Very strong >100; Extreme
When is the posterior belief and the bayes factor the same?
The posterior belief and the Bayes factor are the same if the prior belief is that the distribution is 50/50.
What does statistical evidence refer to?
Statistical evidence refers to a change in conviction concerning a hypothesis brought about by the data.
It is easier to detect the _____ of something than the _____ of something
Presence; Absence
What does a bayes factor smaller than 1 mean?
A Bayes factor small than 1, provides evidence for the null hypothesis over the alternative hypothesis.
What does a bayes factor of approximately 1 indicate?
that the experiment was not sensitive enough to differentiate between the two hypotheses. This is how power is incorporated into Bayesian statistics.
What does the liklihood principle state?
The likelihood principle states that all the information
relevant to inference contained in data is provided by the likelihood.
What does a hypothesis having the highest likelihood mean?
A hypothesis having the highest likelihood does not mean that the hypothesis has the highest probability of being true, it means that the data support the hypothesis the most.
Distinguish between the p-value and likelihood in relation to a graph
In a distribution, the p-value is the area under the curve, whereas the likelihood is the height of the distribution at a certain point.
Give three advantages to the Bayes factor
- The Bayes factor provides a continuous degree of evidence without requiring an all-or-none decision (p-value).
- The Bayes factor allows evidence to be monitored during data collection.
- The Bayes factor differentiates between support for the null hypothesis (evidence for absence of an effect) and non-informative data (absence of evidence).
Name 4 advantages to frequentist statistics
- The p-value is objective as the probability of the data given the null hypothesis (P(data | hypothesis) is an objective probability.
- Frequentist statistics also allows to control for Type I and Type II error rates.
- Frequentist statistics are very practical as almost all research designs can use null hypothesis
testing. - The p-value always has the same interpretation.
Comment on the view of frequentists on the probability of a single event.
The probability of a single event (e.g. P(hypothesis)) does not exist according to frequentists.
Name four practical advantages to bayesian statistics
- Bayesian statistics allows for learning from prediction errors.
- Bayesian statistics allows quantifying evidence in favour of a hypothesis.
- Bayesian statistics allows adjusting knowledge while conducting research.
- Bayesian statistics allows to obtain answers to meaningful questions (e.g. P(hypothesis|data)).
What are the three types of distributions in bayesian statistics?
- Uniform distribution
This is a distribution where every value is equally likely. - Normal distribution
This is a distribution where one value is most likely with the values on both sides of this value being equally likely as the distribution is symmetrical. - Half-normal distribution
This is a distribution which is centred on zero with only one tail (e.g. positive or negative).