Term
|
Definition
Science is a method or process which tries to describe, explain, predict and prescribe (in a limited sense) empirical phenomena. It must have three different features:
1. logic, hence it must follow the rules of logic
2. objectivity, the method must arrive to the same results independently of the researcher
3. Applications of the method of science should be systematically documented.
Science is a methodology or process, not a specific body of knowledge. |
|
|
Term
What are the assumptions of science? |
|
Definition
1) Nature is orderly and regular
2) Nature is knowable (humans are part of this natural world and are capable of knowing nature and predicting it)
3) Causation – Every event has a natural cause, even if it is not observable
4) Empiricism - Knowledge is based on our experience
5) Nothing is self-evident |
|
|
Term
|
Definition
1) Nature is orderly and regular
2) Nature is knowable
3) Causation – Every event has a natural cause, 4) Empiricism - Knowledge is based on our experience
5) Nothing is self-evident |
|
|
Term
|
Definition
Rationalism is the understanding that all knowledge can be obtained by strict adherence to the forms and rules of logic. |
|
|
Term
|
Definition
| Knowledge based on our perceptions, experience, and observations. |
|
|
Term
|
Definition
| Affirm antecedent (if) or deny consequent (then) |
|
|
Term
|
Definition
| Deny antecedent (if) or Affirm consequent (then) |
|
|
Term
|
Definition
| Begins with a general premise and theory. From that theory, hypotheses are developed to test that theory. The conclusion is very specific and includes what is in the premise. Most powerful type of scientific explanation |
|
|
Term
|
Definition
| Begins with a specific premise and moves to general conclusion. Begins with data collection and theory is then constructed. The conclusion made is general, and goes beyond what we know. Used primarily in social sciences. |
|
|
Term
|
Definition
| Measurement is the assignment of numerals/numbers to objects, events, or variables according to a specific set of rules. There are four levels of measurement: |
|
|
Term
| Nominal (level of measurement) |
|
Definition
| Creates exhaustive and mutually exclusive categories. The categories can not be put into order; no mathematical relationships. Examples: gener, nationality, religion |
|
|
Term
| Ordinal (level of measurement) |
|
Definition
| Categories that can be put into order of value (< or >), but no specific numbers are known (can't measure distance between them). Example: educational levels, degree of conservatism |
|
|
Term
| Interval (level of measurement) |
|
Definition
| Exact distance between each of the observations is composed of fixed and equal units; is isomorphic to the number line. You cannot use it to calculate ratios, because there is no set zero point, and there may be negative numbers. Example: temperatures in different cities, increase or decrease of home value. |
|
|
Term
| Ratio (level of measurement) |
|
Definition
| Isomorphic to the positive number line. There are no negative numbers and a fixed natural zero, which means that ratios can be determined. Example: age (if specific), weight, time, length |
|
|
Term
| Define Measurement Validity |
|
Definition
| Measurement validity is concerned with the question “Am I measuring what I intend to measure?”. There are 4 different types of measurement validity. |
|
|
Term
|
Definition
| Does the measurement look correct? Common sense assessment. |
|
|
Term
|
Definition
| Does the measure cover the universe of the underlying concept? |
|
|
Term
|
Definition
| Can you use the measurement to accurately predict another measure? Should be a strong relationship between the results it predicts and the results it obtains when measuring the same or related variables. |
|
|
Term
|
Definition
| Do the measurements relate the way theory predicts? Does the measure behave like the theory says a measure of that construct should behave? |
|
|
Term
| Define Measurement Reliability |
|
Definition
| When our measures yield consistent and stable results. There are three types of measurement reliability. |
|
|
Term
|
Definition
| Will different testers have different results? If the measurement is valid, the results should be the same. |
|
|
Term
|
Definition
| If you measure the same thing twice with the same instrument, do you get the same result? |
|
|
Term
|
Definition
| Are there any outside factors that affect the measurement? |
|
|
Term
|
Definition
| A hypothesis is a tentative answer to a research problem, expressed in the form of a clearly stated relation between independent and dependent variables. |
|
|
Term
|
Definition
| Represented by X, are used to predict and/or explain why. The one that changes. |
|
|
Term
|
Definition
Represented by Y, are the variables we try to explain.
(y=f(x)) |
|
|
Term
|
Definition
| Represented by Z, are used to test for outside factors that may influence the correlation between X and Y. |
|
|
Term
| What is a sample statistic? |
|
Definition
| A sample statistic is the counterpart of the population parameter in the sample (subset of the population); an estimate of the population parameter |
|
|
Term
| What is a population parameter? |
|
Definition
| A fixed empirical value for the entire population. It is always unknown. A population parameter is an attribute found in the population (complete set of relevant units of analysis) that can be measured |
|
|
Term
| What is a sample statistic and population parameter, and what is the relationship between them? |
|
Definition
| A sample statistic is an estimate of the population parameter. The two are related in that the sample statistic is trying to find an estimate that is as close as possible to the population parameter. The population parameter has one true value, but since it is unknown we use a sample statistic. |
|
|
Term
| What is a probability sample? |
|
Definition
| A probability sample is a sample where we can specify the chance of selecting any element of the population. All units of a population have the same probability of being included in the sample. |
|
|
Term
| Why is it important to have a probability sample? |
|
Definition
| Only probability sampling can be used to estimate the population’s parameters on the basis of the calculated sample statistics; needed to calculate an error margin and make inferences about a population. |
|
|
Term
| Simple random sample (Type of probability sample) |
|
Definition
Good for small populations; based on a simple random selection of units from the sampling frame. Drawing from a hat is an example.
|
|
|
Term
| Systematic random sample (Type of probability sample) |
|
Definition
| You begin with a list, pick a random starting point, and pick every kth unit. There is a problem of periodicity in this method because if the list is not completely random, the selections will not be representative of the population. |
|
|
Term
| Stratified sample (Type of probability sample) |
|
Definition
| Uses information already known about the population to strategically pick samples. It only uses characteristics with known percentages. |
|
|
Term
| Cluster sample (Type of probability sample) |
|
Definition
| Randomly select a cluster of units. Then, randomly select units from the individual clusters. The clusters are selected by simple random sampling or by stratified sampling. |
|
|
Term
| Determinants of sample size (3) |
|
Definition
Practicality – How much time and money is there to conduct data collection?
Math – The results should yield an acceptable error margin and produce sophistication of data analysis. The larger the sample, the lower the error margin and the more specifics can be drawn.
Population characteristics – The more homogenous the population is, the smaller the sample size needs to be. |
|
|
Term
| What is a non-probability sample? |
|
Definition
| There is no way of specifying the probability of each unit’s inclusion in the sample. Inferences to the population can’t be made with these kind of samples. |
|
|
Term
| Quota Sample (Type of Non-Probability Sample) |
|
Definition
| Select samples to fill certain quotas such as gender, age and place of residence. Not commonly used. |
|
|
Term
| Purposive/Judgmental Sample (Type of Non-Probability Sample) |
|
Definition
| Use subjective judgment and attempt to select sampling units that appear to be representative of the population. An example of this sample is a focus group. |
|
|
Term
| Convenience/Available Sample (Type of Non-Probability Sample) |
|
Definition
| Relies on convenient/available subjects. Examples are polls from radio stations and Internet surveys. |
|
|
Term
| Snowball Sample (Type of Non-Probability Sample) |
|
Definition
| Used to reach hard-to-reach populations. When you can’t compile a sampling frame, you try to find one person who fits your criteria and ask them to refer you to other subjects. Used by journalists. |
|
|
Term
| Internal (Causal) Validity |
|
Definition
Does X cause Y? To be internally valid, we must show:
a) Correlation (X can’t cause Y if they do not move together)
b) time order (X has to come before Y if it is the cause of Y)
c) non-spuriousness (X and only X causes Y) |
|
|
Term
|
Definition
External validity concerns itself with the question of generalizability. If an experiment is externally valid, we should be able to generalize or imply things about other populations and settings. To be externally valid, we must have:
a) a randomly selected probability sample
b) a “real world” setting |
|
|
Term
| What is a classic experimental design? |
|
Definition
The classic experimental design has two groups, an experimental group and a control group. The experimental group is given a stimulus, whereas the control group is not. Both groups are tested both before and after the stimulus. Afterwards, researchers compare the results of both groups’ tests.
Pretest Stimulus Posttest
Experimental Group T¹¹ X T¹²
Control Group T²¹ T²² |
|
|
Term
| Discuss the internal and external validity of classic experiments and the factors that impact internal and external validity of an experimental design. |
|
Definition
| The classic experimental design can be used to show internal validity if the groups are equivalent to begin with. Since the researcher controls the data, it is fairly straightforward to show correlation and time order. Non-spuriousness is also provable, but it is more difficult, since we must prove that only X causes Y. The classic experimental design is externally valid if the samples for the groups are randomly selected and the experiment takes place in a real world setting. |
|
|
Term
| Extrinsic Factors that impact internal validity of experimental design. |
|
Definition
| Extrinsic factors: ethical considerations and issues of practicality sometimes prevent the random assignment of research participants. But random assignment yields to equivalent groups, so as long as the groups are randomly assigned the extrinsic factors are controlled. |
|
|
Term
| Intrinsic Factors that impact internal validity of experimental design. |
|
Definition
2) Intrinsic factors: include changes that occur during the study period.
a) History
b) Experimental mortality
c) Maturation
d) Instrumentation
e) Testing reactivity |
|
|
Term
| Classic Experimental Designs are strong and weak in what? |
|
Definition
| Classic Experimental designs are strong in internal validity but weak in external validity. |
|
|
Term
| What is a correlational design? |
|
Definition
| Correlational designs collect data at one moment of time. This design is often identified with survey research. A correlational design looks at a property/disposition relationship, and is used to describe a statistical association between the two. It allows us to generalize. It takes place in a real world setting (population surveys, polls, etc.). |
|
|
Term
| Difference Between Correlational and Experimental Designs? |
|
Definition
| In correlational, studies property-disposition relationships whereas in experimental designs the relationship studied is of the stimulus-response type. Differing from the experimental design, correlational designs only use one group of units and measures are only taken at one time moment.Unlike the classic experimental design, the correlational design uses real world settings. Since researchers can not manipulate the data as easily as the experimental design, it is never possible to prove non-spurriousness. |
|
|
Term
| Correlation Designs are strong and weak in what? |
|
Definition
| Correlational designs have weak internal validity and strong external validity. |
|
|
Term
| Discuss the internal and external validity of correlational designs. |
|
Definition
| Correlational designs have strong external validity and generalizibility because they use a randomly selected probability sample and data is collected in a “real world” environment. However, the design is weak on proving internal validity, because you can never prove non-spuriousness. Correlation is usually straightforward, and time order may or may not be provable, but non-spuriousness is always a problem. Also often impossible to get true random probability sample. |
|
|
Term
| What are quasi-experimental designs? |
|
Definition
| A quasi-experimental design has two or more groups and measures over time. It may or may not use a probability sample, and it may or may not have stimuli. It is used in public policy research to look at program impacts, policy effects, etc. |
|
|
Term
| Quasi-experimental designs are strong and weak in what? |
|
Definition
| Have weaker internal validity than experimental designs but stronger internal validity than a correlational design. External validity depends on if probablility sample exists. |
|
|
Term
| Discuss internal and external validity of quasi-experimental designs. |
|
Definition
| In quasi-experimental designs, the internal validity is weaker than a true experiment because the data cannot be manipulated. While internal validity is stronger than a correlational design, it is not as strong as a true experiment. The external validity varies. If there is a probability sample and the setting replicates the real world, the external validity is strong. |
|
|
Term
| Discuss the criteria for inferring causation. |
|
Definition
The demonstration of causality involves three distinct operations:
a) Correlation or covariation: two or more phenomena vary together.
b) Time order: requires demonstrating that the assumed cause occurs first or changes prior to the assumed effect.
c) Nonspuriousness: demonstrate that the relation between two variables cannot be explained by a third variable. |
|
|
Term
|
Definition
| Nomothetic explanations look at a sample, and try to observe probable relations between X and Y. |
|
|
Term
|
Definition
| Ideographic explanations use narratives, since it is impossible to draw a sample, and there is only one case - like why the Berlin Wall fell. |
|
|
Term
| Random Assignment DOES NOT EQUAL RANDOM SELECTION (for external validity) |
|
Definition
| Random Assignment DOES NOT EQUAL RANDOM SELECTION (for external validity) |
|
|
Term
| What is a cross-sectional design? |
|
Definition
| Cross-Sectional design looks at data collected at one point in time. It observes the population at one time, which enables the researcher to compare independent variables. Sometimes used to detect prevalence of disease. Same as correlational. |
|
|
Term
| What is a longitudinal design? |
|
Definition
| A longitudinal design involves repeated series of measurements over a period of time. Panels and time-series studies use this to see changes in the dependent variable. |
|
|
Term
| What are measures of central tendency? |
|
Definition
| Measures of central tendency are statistical measures that reflect a typical or average characteristic of frequency distribution (mode, median, mean) |
|
|
Term
|
Definition
| Dispersion is the extent of distance from the central value. It is used in order to observe the variations in a sample (proportion outside the mode, range measures, variance, standard deviation) |
|
|
Term
|
Definition
a) Overall range: distance between the highest and lowest values of the distribution.
b) Interquartile range: difference between the lower and upper quartiles, i.e. from the 25% cumulative percentile to the 75% cumulative percentile.
c) Interdecile range: the difference between the first and the ninth deciles (10% and 90%). |
|
|
Term
|
Definition
| the average of the squared deviations from the mean. |
|
|
Term
|
Definition
| Equal to the square root of the variance. Unlike variance, it expresses dispersion in the original units of measurement. |
|
|
Term
|
Definition
Logic
Objectivity
Systematic Documentation |
|
|
Term
| Types of Measurement Reliability |
|
Definition
a) Instrument reliability
b) Observer reliability
c) Phenomenon reliability |
|
|
Term
| Four Types of Probability Samples |
|
Definition
a) Simple random sample
b) Systematic samples
c) Stratified samples
d) Cluster samples |
|
|
Term
| Types of Measurement Validity |
|
Definition
1) Face
2) Content
3) Predictive
4) Construct |
|
|
Term
|
Definition
| Science is a methodology or process, not a specific body of knowledge. |
|
|
Term
|
Definition
1) Nominal
2) Ordinal
3) Interval
4) Ratio |
|
|
Term
| Types of NON-Probability Samples |
|
Definition
1) Quota
2) Purposeful/Judgmental
3) Available
4) Snowball |
|
|
