1/77
Looks like no tags are added yet.
What are the four types of validity used to evaluate research quality?
Construct validity, internal validity, external validity, and statistical validity. Each examines a different aspect of the quality of a research conclusion.
Construct validity
The extent to which the variables and measurement instruments accurately represent the theoretical concepts researchers intend to study.
Internal validity
The extent to which researchers can conclude that one variable causes changes in another rather than the relationship being explained by alternative factors.
External validity
The extent to which research results can be generalized beyond the particular individuals or circumstances studied to the intended population or other settings.
Statistical validity
The extent to which statistical conclusions about relationships between variables are appropriate, justified, and supported by suitable statistical methods and evidence.
What is the difference between construct, internal, external, and statistical validity?
Construct validity concerns measurement; internal validity concerns causal conclusions; external validity concerns generalization; statistical validity concerns the correctness and strength of statistical conclusions.
three types of claims
frequncy claim
assosiation claim
causal claim
Frequency claim
A claim describing the frequency or level of a variable in a particular population, such as the percentage of students experiencing stress.
Association claim
A claim stating that two measured variables are related, such as greater gratitude being associated with higher subjective happiness.
Causal claim
A claim stating that one variable produces a change in another, such as an intervention causing an improvement in happiness.
What is the difference between a frequency claim, association claim, and causal claim?
A frequency claim describes one variable; an association claim concerns the relationship between variables; a causal claim asserts that one variable causes changes in another.
Which type of validity evaluates whether gratitude and happiness were accurately measured?
Construct validity, because it concerns whether the instruments adequately measure gratitude and happiness.
Which type of validity evaluates whether findings from psychology students can be generalized to other adults?
External validity, because the question concerns applying findings beyond the participants actually studied.
Which type of validity evaluates whether gratitude actually causes happiness rather than merely correlating with it?
Internal validity, because the question concerns causal inference and alternative explanations.
Which type of validity evaluates whether a correlation test was appropriate and whether its conclusions are justified?
Statistical validity, because it concerns the suitability of statistical methods and interpretation of results.
APA reporting - correlations
needed: n, r, p-value
n=30, r = -.417, p=.022
r(28) = -.417, p=.022
r(n-2)
Why is internal validity not always the main priority in correlational research?
If the goal is to describe an association rather than establish a causal effect, researchers primarily need to evaluate whether the association was measured and analyzed correctly.
What information in a scientific article is especially useful for assessing external validity?
The Methods section, particularly the Participants subsection, because it explains who participated, how they were recruited, and the population to which the findings may apply.
Why can recruiting only undergraduate psychology students threaten external validity?
The participants may differ systematically from the wider population, making it difficult to generalize the findings to people of different ages, educational backgrounds, or experiences.
What is the importance of sample selection for external validity?
A sample that adequately represents the target population allows stronger generalization than a sample selected only for convenience.
How should a Pearson correlation be reported in APA style?
Using the format r(df) = correlation coefficient, p = p-value, together with a sentence describing the relationship and its direction.
What does a negative correlation of r = −.417 between stress and impulsivity mean?
Higher stress scores tend to be associated with lower impulsivity scores, and lower stress scores with higher impulsivity scores. This does not establish causation.
What does p = .022 indicate for a correlation between stress and impulsivity at α = .05?
The relationship is statistically significant because .022 < .05, so the null hypothesis of zero population correlation is rejected.
What is the difference between r and p when reporting a correlation?
r describes the direction and strength of the observed correlation; p indicates how compatible the result is with the null hypothesis under the chosen statistical test.
What is the relationship between gratitude and subjective happiness when r = .276 and p < .05?
There is a positive, statistically significant association: higher gratitude scores tend to accompany higher subjective happiness scores.
What does a correlation of r = −.288 between negative affect and subjective happiness indicate?
A negative relationship: higher negative affect tends to be associated with lower subjective happiness.
Does a statistically significant correlation between kindness and subjective happiness prove that kindness causes happiness?
No. Statistical significance supports the existence of an association but does not establish causation.
How can a researcher assess the statistical validity of a correlation result?
Evaluate statistical significance using the p-value, test suitability using assumptions and correct execution, relevance using effect size, and accuracy using a confidence interval.
What are the four components of statistical validity assessment?
Significance (p value)
suitability of the statistical test (checking the assumptions and execution of the test)
relevance (effect size)
accuracy (confidence interval)
Suitability of a statistical test
1. Check assumptions
2. Check if hypotheses match expectations
3. Check if results match hypothese
What are the three assumptions for using Pearson's correlation? what if they’re broken?
The sample is random → if not lowex external validity and low laidity of mathematical results
both variables are measured at the interval or ratio level → if one or both are ordinal chi-square test
the relationship between variables is linear → non linear? : check the sape → if monotone Spearman corelation coefficient
Random sampling assumption for Pearson correlation
The sample should be randomly selected so that statistical results and their generalization to the intended population are more justifiable.
Interval/ratio measurement assumption for Pearson correlation
Both variables should be quantitative variables measured at the interval or ratio level rather than nominal or ordinal categories.
Linearity assumption for Pearson correlation
The relationship between the two variables should be approximately linear, meaning it can be reasonably represented by a straight-line trend.
Why is the linearity assumption important for Pearson's r?
Pearson's r describes linear associations. A strong but nonlinear relationship may receive a misleadingly low Pearson correlation.
What happens if a correlation study uses a nonrandom sample?
External validity is reduced, and the inferential conclusions may be less reliable for the target population because selection may create systematic differences.
What should researchers do when both variables are nominal categorical variables?
Use a statistical method designed for categorical data, such as the chi-squared test of independence, rather than Pearson's correlation.
Spearman correlation coefficient (rₛ)
A rank-based measure of the direction and strength of a monotonic relationship, suitable for ordinal variables and certain nonlinear relationships (only increasing/decreasing, not both)
Pearson's r versus Spearman's rₛ
Pearson measures linear association using the original quantitative values; Spearman measures rank association, allowing analysis of monotonic relationships that may be nonlinear.
When is Spearman's correlation preferred over Pearson's correlation?
When variables are ordinal, when the relationship is monotonic but nonlinear, or when influential outliers substantially distort Pearson's correlation.
Monotonic relationship
A relationship in which one variable consistently increases or consistently decreases as the other increases, even if the relationship is curved.
Nonmonotonic relationship
A relationship that changes direction, increasing over part of the range and decreasing over another part, or vice versa.
Can Spearman's correlation be used for a curved relationship that always increases?
Yes. The relationship is monotonic increasing, so a rank-based correlation can describe the association even though the relationship is not linear.
Can Spearman's correlation adequately describe a U-shaped relationship?
No. A U-shaped relationship decreases and then increases, making it nonmonotonic. A single Spearman correlation cannot adequately represent that pattern.
Which correlation should be used for ordinal rankings of fairness and forgiveness?
Spearman's rank correlation, because both variables can be represented by ordered ranks.
Why can outliers be problematic for Pearson's correlation?
Pearson's correlation is sensitive to extreme observations, which may substantially increase or decrease the estimated linear relationship.
What should researchers do before removing or changing an outlier?
Investigate whether the observation reflects an error. Correct documented errors, but do not remove valid observations simply because they are extreme.
What should researchers do if an outlier is valid but strongly changes Pearson's correlation?
Compare Pearson's r with and without the outlier; if the difference is substantial, the course recommends using Spearman's correlation.
What does it mean if Pearson's r is .706 with an outlier and .913 without it?
The outlier substantially weakens the apparent linear association, showing that Pearson's correlation is sensitive to this observation.
Why might Spearman's correlation be less affected by an outlier?
Spearman uses ranks rather than the original numerical distances between values, reducing sensitivity to extremely large or small values.
How does Spearman's correlation work?
Replace the original observations with ranks and calculate the correlation between those rank scores using the Pearson correlation formula.
Rank score
The numerical position of an observation when values are ordered from smallest to largest.
How are rank scores assigned when calculating Spearman's correlation?
Sort each variable's values from smallest to largest, assign ranks beginning with 1, and calculate the correlation using those ranks.
In a forgiveness dataset with sorted values 2.2, 2.8, 3.0, and 3.4, what are the rank scores?
The ranks are 1, 2, 3, and 4 respectively, because they are ordered from smallest to largest.
In a forgiveness dataset where 4.1 is the sixth-smallest value, what rank does 4.1 receive?
Rank 6.
Why can rank transformation make a nonlinear monotonic relationship more suitable for correlation analysis?
Rank transformations preserve the ordering of observations and may transform a curved monotonic pattern into an approximately linear pattern between ranks.
What is the statistical notation for a Spearman rank correlation?
rₛ for the sample rank correlation and ρₛ for the corresponding population rank correlation.
What hypotheses are used for a two-sided Spearman correlation test between fairness and forgiveness?
H₀: ρₛ = 0; Hₐ: ρₛ ≠ 0. The hypotheses test whether the population rank correlation differs from zero.
What does Spearman rₛ = .902 with p < .001 indicate for fairness and forgiveness?
A very strong positive rank association that is statistically significant, providing evidence against the null hypothesis of zero population rank correlation.
How should a Spearman correlation of rₛ = .902, p < .001 between fairness and forgiveness be reported?
There is a significant positive rank correlation between fairness and forgiveness scores, rₛ = .902, p < .001.
How does Spearman rₛ = .831 compare with Pearson r = .706 in a dataset containing an outlier?
Both suggest a positive association, but the higher Spearman coefficient indicates that the rank relationship is less weakened by the extreme observation.
What is a directional (one-sided) hypothesis?
A hypothesis specifying the expected direction of a relationship, such as Hₐ: ρ > 0 for a predicted positive correlation.
What is a nondirectional (two-sided) hypothesis?
A hypothesis predicting that a relationship exists without specifying its direction, expressed as Hₐ: ρ ≠ 0.
What is the difference between one-sided and two-sided correlation tests?
A one-sided test evaluates evidence in a specified direction; a two-sided test evaluates evidence for either positive or negative correlations.
What is the advantage of a one-sided hypothesis test?
It directly matches a directional theoretical prediction and concentrates the test on the expected direction of the relationship.
What is the disadvantage of a one-sided hypothesis test?
If the observed relationship is in the opposite direction, it does not support the directional alternative, even if the result looks statistically strong.
What is the advantage of a two-sided correlation test?
It allows researchers to detect statistically significant evidence of a relationship in either direction.
What is a disadvantage of a two-sided test compared with a justified one-sided test?
does not match the expectations of the researcher
less likely to adopt a new theory
Why must researchers check the sign of a correlation before interpreting a directional hypothesis test?
A statistically significant finding in the opposite direction does not support a research hypothesis predicting a specific positive or negative relationship.
A researcher predicts that greater financial means are associated with higher happiness, but finds r = −.132. Does the result support the hypothesis?
No. The observed relationship is negative, whereas the hypothesis predicts a positive correlation.
A researcher tests H₀: ρ = 0 versus Hₐ: ρ > 0 at α = .05 and observes r = −.132. Can they reject H₀ in favor of the positive alternative?
No. The observed correlation is in the opposite direction from the specified alternative.
Why is a small p-value insufficient if the observed correlation contradicts the researcher's directional hypothesis?
The p-value must be interpreted in relation to the actual test and hypothesis. Evidence for a relationship in the opposite direction cannot establish the originally predicted directional effect.
Why should statistical hypotheses be formulated before inspecting the observed correlation?
To ensure the test reflects the researcher's theoretical expectations rather than choosing a favorable direction after seeing the data.
What are the three major steps when choosing a correlation method?
Check the measurement levels of the variables, examine the relationship's shape, and investigate whether outliers substantially influence the statistic.
A researcher measures income and happiness on quantitative scales and observes an approximately straight upward trend. Which correlation is suitable?
Pearson's correlation, provided its sampling and measurement assumptions are sufficiently satisfied.
A researcher measures ranked job satisfaction and ranked motivation. Which correlation is suitable?
Spearman's rank correlation, because the data are ordinal.
A researcher observes a curved but consistently decreasing relationship between stress and performance. Which correlation is suitable?
Spearman's rank correlation can describe the monotonic decreasing association.
A researcher observes an inverted U-shaped relationship between arousal and performance. Is Spearman's correlation sufficient?
No. The relationship increases and then decreases, making it nonmonotonic; a different analysis is needed to represent the pattern adequately.