SOM/Ethics
What is the difference between an independent variable vs. independent variable vs. confounding variable?
Independent Variable: predictor variable as its valuable does not depend on any other variable
Dependent Variable: outcome variable as the value does depend on another variable ( e.g independent variable)
Confounding Variable: A variable that affects both the dependent variable and independent variables and causes distortion of the effects of the study.
What are the 3 characteristics of a correlation?
The Direction of the Relationship
The Form of the Relationship
The Degree of the Relationship
What makes any measure or task a good way to measure the construct you wish to study?
Reliability & validity
** Remember reliability is needed for validity but validity is NOT needed for reliability
Sensitivity
Correctly measuring how many people have the particular condition
A / (A+C) = True Positive / Disease Present
Column for disease present on the table slide 4 lecture 5
Central limit theorem
For a relatively large sample, the sample statistics (x bar) is approximately normal distributed, regardless of the population distribution. Please note that this will hold where: If distribution is already normally distributed and/or if the sample size is large (e.g larger than 30) but the distribution pattern is not know
What was the outcome of the Tuskegee Study?
The birth of the National Research Act ( 1974) which created the National Commission for the Protection of Human Subjects of Biomedical and Behavioral Research --> The Belmont Report in 1979 which is aimed at protecting human subjects during scientific research
What type of variables does a correlation use to describe a relationship?
Two continuous variables
What is the difference between generalizability theory vs classical test theory vs item response theory?
Generalizability theory --> multiple sources of error
Classical test theory --> error is random
Item response theory --> helps us evaluate test takers performance to specific test items to understand latent traits
Specificity
Correctly identifying how many people who are testing negative given that they do not have the disease
D/(B+D)=(True Negative)/Disease not Present
Column for disease not present slide 4 lecture 5
Null vs alternative hypothesis?
If we reject the Null Hypothesis (Ho): indicates that there is significant evidence for the alternative hypothesis
If we fail to reject the null hypothesis (Ho): indicates that there is not a significant amount of evidence for the alternative hypothesis
What is the purpose of an IRB?
Purpose of the IRB is to support researchers identify potential harms to study participants and help to assess the benefits compared to the risks
What are the 3 measures of central tendency?
Mean, Median, Mode
Different types of reliability measures?
Temporal --> test retest, parallel tests
Internal --> split half, KR coefficient, cronbach's alpha
Across administrators --> interrater, kappa coefficient
NPV & PPV
Negative predictive value: Probability that I test negative and I do not have the disease: D/(D+C) - row for test negative on table
Positive predictive value: Probability that I test positive and I actually have the disease:A/(A+B) - row for test positive on the table
Type I vs Type II error?
Type I error: A statistical decision-making error in which a large amount of sampling error causes the rejection of the null hypothesis when the null hypothesis is true. (saying there is a relationship when there is NOT)
Type II error: A statistical decision-making error in which a large amount of sampling error causes the acceptance of the null hypothesis when the null hypothesis is false. (saying there is no relationship when in fact it DOES exist)
What is the difference between descriptive vs. inferential statistics?
Descriptive Statistics: Procedures for organizing and summarizing data so that the important characteristics of the data can be described and communicated.
Inferential Statistics: Procedures for determining whether the sample data are representative of a particular relationship in the population or allow you to draw conclusions about your data that can be applied to the wider population.
What are the threats to internal validity in experimental research?
1. History
2. Maturation
3. Testing threat
4. Instrumentation threat
5. Mortality / attrition
6. Statistical regression to the mean
Tripartite model for validity?
Content validity --> accuracy with which an item provides information about construct being measured
Construct validity --> examination of how accurately & usefully a test measures the specified constructs (convergent vs discriminant)
Criterion validity --> can either measure what is going on currently (concurrent) or provide a prediction of what will happen in the future (predictive)
** Also know face validity --> occurs when the test takers and the test user believe the test is accurately measuring what is intended to
Receiving operating characteristic (ROC) curve
References to the performance across all thresholds. Obtained by plotting all sensitivity values ( true positive fraction) on the y axis against their equivalent ( 1-specificity) values or false positive values for all available thresholds on the x axis.
**Used to evaluate accuracy of diagnostic test & compare to others to determine which one is better
What do a, 1-a, 1-B, & B represent in terms of hypothesis testing?
a = type I error (reject Ho when Ho is true)
1 - a = correct decision (accept Ho when Ho is true)
B = type II error (accept Ho when Ho is false)
1 - B = correct decision (reject Ho when Ho is false)
What are the five scales of measurement and what does each one consist of?
Nominal Scale: Observations are labeled and categorized
Ordinal Scale: Ranking observations in terms of size or magnitudes.
Interval Scale: EQUAL differences (or intervals) between numbers on the scale reflecting differences in magnitude. (Ratios of magnitudes are not meaningful).
Ratio Scale: Ratio of numbers that do reflect ratio of magnitudes. This scale has an absolute zero point.
What is the difference between the following study designs: experimental, correlation, quasi-experimental, vs. observational?
Experimental Design: A research procedure in which one independent variable is actively changed or manipulated, the scores on another variable (dependent variable) are measured, and all other variables are held constant to determine whether there is a relationship. It can also determine a cause-and-effect relationship between two variables
Correlation Design: A research procedure in which subjects' scores on two variables are measure, without manipulation of either variable, to determine whether there is a relationship. It cannot determine a cause-and-effect relationship
Quasi-Experimental: A research procedure in which the researcher does not directly manipulate the independent variable and/or the researcher does not randomly assign subjects to treatment conditions. It examines differences between preexisting groups of subjects or differences between preexisting conditions (aka subject variables). The variable that is used to differentiate the groups is called the quasi-independent variable, and the score obtained for each individual is the dependent variable
Observational Designs: a research method where a researcher observes and records data about individuals or phenomena without interfering or manipulating any variables
What is MultTrait-Multimethod Matrix (MTMM) used for?
Reliability Diagonal: ( Monotrait-Monomethod): This tells you the correlation of reliability for each of the Traits with the methods described (HIGHEST CORRELATION)
Validity Diagonals: (Monotrait-Heteromethod): Correlations between measures of the same trait measures using different methods. Two measures are of the same trait or concept, we would expect them to be strongly correlated (CONVERGENT VALIDITY)
Heterotrait-Monomethod Triangles: These are the correlation among different traits that share the same method of measurements. If these are high, it is due to having a strong methods factor (DISCRIMINANT VALIDITY)
Heterotrait-Heteromethod Triangles: These are correlations that differ in both trait and method. We expect these to be the lowest in the matrix (further illustrates DISCRIMINANT VALIDITY)
Area under the curve (AUC)
The area under the ROC curve provides measure of overall accuracy of a diagnostic test. AUC = 1 = perfect accuracy.
What are the differences between within subjects, between subjects, vs matched subjects design?
Within subjects design --> repeated measures - same subjects in all conditions (more power)
Between subjects design --> subjects are only in one treatment condition
Matched subjects design --> matching subjects on one or more characteristics assumed to be relevant to behavior in the study
**Know pros/cons for each one