EPPP Research

Question 1

-Qualitative Research

Answer 1

= obtain a description of the quality of relationships, actions, situations, or other phenomena. Naturalistic, contextual approach emphasizing understanding and interpretation and primarily inductive. Investigator’s perspective important. Observation, interviews and document analysis are strategies.

Question 2

Quantitative Research

Answer 2

= obtain numerical data on variables. Empirical methods and statistical procedures, emphasizes prediction, generalizability, causality and id deductive. Investigators perspective minimized. Can be non-experimental (descriptive – to collect data on variables rather than to test hypotheses about relationships between them, ie correlational) or experimental (conducted to test hypothesis about effects of 1 or more IV or DV).

Question 3

Planning and Conducting Research (steps)

Answer 3

• Developing an Idea into a Testable Hypothesis about the relationship between the variables. • Choosing An Appropriate Research Design. • Selecting a Sample. ID the target population and detmine how the sample will be selected from pop. • Conducting the Study and collect and record data. • Analyzing the Obtained Data using appropriate descriptive and inferential statistical techniques. • Reporting the Results

Question 4

Independent Variable

Answer 4

(treatment or intervention, must have at least 2 levels = X): When it affects or alters status of another variable = Dependent Variable (outcome which is observed and measured = Y). “What is the effect of the Independent variable on the dependent variable?”

Question 5

Defining and Measuring Variables:

Answer 5

Once IV and DV ID’d, they must be operationally defined in terms of method or process used to ID or measure them. Sometimes it is a behavior to be observed and must decide how to record or measure it. Protocol Analysis is a form of content analysis, verbalizations thought out loud of problem solving are recorded and coded in terms of relevant categories like intentions, cognitions, planning and evaluations.Can look at a sample rather than complete record. Interval recording useful to divide complex behaviors up or when there is not clear beginning or end. Event sampling, observe behavior each time it occurs, good if it is infrequent or long or that leave a permanent record, ie worksheet. Situational sampling observing in a number of settings. Sequential analysis, coding sequences to study complex social behaviors.

Question 6

True experimental research

Answer 6

provides enough control to conclude that observed variability in DV is caused by variability in IV, through control of conditions, levels of variables and *****random assignment of subjects to groups.

Question 7

Quasi-experimental research

Answer 7

= can not control assignment of subjects to groups and must use intact pre-existing groups or single treatment group.

Question 8

Sampling Techniques

Answer 8

Sample must be representative of population to maximize generalizability of results. Systematic sampling (selection) techniques include: • Simple Random Sampling: Every person has = chance of being picked. • Stratified Random Sampling: If population varies in specific “strata”, use stratified random sampling to ensure that each stratum is represented in the sample. Divide according to specific strata and then randomly select subjects from each stratum ie age, race, education. • Cluster Sampling: selecting units of individuals and then select all or randomly select from unit (multistage cluster sampling) . Useful if don’t have access to full population.

Question 9

Methods of Control in Experimental Research

Answer 9

2 questions: Is there a relationship between the IV and DV? If so, is the relationship a causal one? To accurately answer, depends on extent design allows control of 3 factors that cause variability in DV: • The IV (experimental variance) • Systematic error (error due to extraneous variables) • Random error (error due to random fluctuations in subjects, conditions, methods of measurement) To be more certain observed variability in DV is due to IV than error: Choose design that: • Maximizes variability in DV due to IV. • Control variability due to extraneous variables • Minimize variability caused by random error Control is maximized if true experimental study.

Question 10

Maximizing Variability Due to the Independent Variable(s)

Answer 10

True experimental research enhances ability to maximize variability due to IV by allowing levels of IV to be as different as possible so effects on DV can be detected.

Question 11

Extraneous (confounding) variable

Answer 11

is a source of systematic error, irrelevant to he purpose of the study but confounds results as it has a systematic effect on (correlates with) the dependent variable.

Question 12

Random Assignment of Subjects to Treatment Groups (Randomization):

Answer 12

equalizes all effects and is the most “powerful” method of experimental control which makes it true experimental research.

Question 13

Holding the Extraneous Variable Constant:

Answer 13

Eliminate effects by selecting subjects who are homogenous with respect to the variable. Problem: Limits generalizability of research results.

Question 14

Matching Subjects on the Extraneous Variable

Answer 14

: Match subjects in status on that variable and randomly assign matched subjects to 1 of the treatment groups. Matching is useful when sample size is too small to guarantee random assignment will equalize groups with regard to the extraneous variable.

Question 15

Building the Extraneous Variable into the Study (“Blocking”):

Answer 15

Control the extraneous variable by including it is the study as an additional IV so effects on DV can be statistically analyzed. Subjects are blocked (grouped) on the basis of their status on the extraneous variable. Subjects in each block are then randomly assigned to 1 of the treatment groups. ie, blocked in terms of severity of an illness.

Question 16

Statistical Control of the Extraneous Variable:

Answer 16

When you know each subjects status on the EV, can use the ANCOVA or other statistical technique to remove variability in the DV which is due to the EV, by equalizing all subjects with regard to their status on the variable to remove the effects. Useful in quasi-experimental research in which can not randomly assign to treatment groups.

Question 17

Minimizing Random Error

Answer 17

Through experimental research, can minimize random fluctuations by controlling conditions and procedures ie make sure subjects do not become fatigued, they are free from distractions and that measuring devices are reliable. Important: In choosing research design, pick one that minimizes the effects of both systematic and random error and understand the differences between them.

Question 18

Threats to Internal Validity

Answer 18

Internal Validity when it allows an investigator to determine if there is a causal relationship between the IV and DV. Internal Validity is threatened when can not control the 3 sources of variability in previous section. Campbell and Stanley: 7 generic extraneous variables that if not controlled can threaten study’s internal validity: Maturation, History, statistical regression, testing, selection, attrition, interaction with selection

Question 19

Maturation IV Threat:

Answer 19

Any biological or psychological change within subjects during the course of a study as a function of time, not relevant to the research hypo and effects the status of subjects on DV in systematic way. ie, fatique, boredom, hunger, intellectual growth. Control by including more than 1 group in the study and randomly assign subjects to each group since they will all be subject to maturation. Single group time series also: measure DV several times at regular intervals before of after the intervention is applied. Does noteliminate but provides information to detect them.

Question 20

History IV Threat:

Answer 20

when an external event systematically affects the status of subjects on the DV. Historical events a problem with 1 group and even occurs about same time as the IV applied. Include more than 1 group and randomly assign subjects. Ensures all equal in exposure to external event.

Question 21

Testing IV Threat:

Answer 21

When pre test affects subjects scores on the posttest. Administer DV only once or use 2 groups.

Question 22

Instrumentation IV Threat:

Answer 22

Changes in measuring devices or procedures such as rater’s increased accuracy. Include more than 1 group and ensure all are subject to the same instrumentation effects.

Question 23

Statistical Regression IV Threat:

Answer 23

Extreme scores “regress”, move toward the mean when the measure is readministered to the same group of people = statistical regression. Happens when selected due to extreme scores. Include more than 1 group and that all groups have some who are similarly extreme.

Question 24

Selection IV Threat:

Answer 24

When method used to assign subjects results in systematic differences between groups at beginning of the study. Problem with intact groups. Control by random assignment or pretest to subjects to determine if they differ initially with the DV.

Question 25

Attrition IV Threat:

Answer 25

When those who drop out are different in an important way. Pretests can determine if drop outs differed.

Question 26

Interactions with Selection IV Threat:

Answer 26

With groups initially nonequivalent, selection can act alone or interact ie with history if 1 group exposed and 1 not. Selection is really an assignment problem.

Question 27

External Validity

Answer 27

when its findings can be generalized to other people, settings and conditions. Can distinguish Population validity (generalize to other people) and ecological validity (generalize to other settings, problem in analogue studies in laboratory or non-naturalistic setting). External validity is always limited by its internal validity, but a high internal validity does not guarantee external validity. Campbell and Stanley ID’d 4 factors that threaten external validity: Interaction Between Testing and Treatment, Interaction between selection and treatment, reactivity, multiple treatment

Question 28

Interaction Between Testing and Treatment Threat to EV:

Answer 28

Pretest can sensitize to purpose of the study and alter reaction to the IV. Control by not using a pre test or using Solomon 4group design which enable investigator to measure the impact of pretesting on both the external and internal validity of the study. Pretest is treated as an additional IV so effects on the DV are statistically analyzed.

Question 29

Interaction Between Selection and Treatment Threat to EV:

Answer 29

When subjects included have characteristics that make them respond to the IV in a particular way, ie they are volunteers who tend to be more motivated than non volunteers so more responsive to the IV. Ensure that the sample is representative of the population of interest.

Question 30

Reactivity (Reactive Arrangements):

Answer 30

Respond in a certain way because subjects know being noticed or observed= reactivity. May have evaluation apprehension. Can also be altered by demand characteristics or cues in the experimental setting that inform subjects of the purpose of the study or suggest what is expected of them. Biased by experimenter expectancy: provide cues to subjects or computational errors are likely to support research hypothesis. Recheck what conflicts with hypothesis than what supports it. Control by deception, unobtrusive (nonreactive)measures or single (don’t know what treatment group assigned) or double blind technique( subjects nor experimenter know which group subjects assigned to).

Question 31

Multiple Treatment Interference (Order Effects, Carryover Effects):

Answer 31

When exposed to 2 or more levels of IV in within subjects design: effects of one level can be effected by effects of previous exposure. Control by counterbalanced design,: different subjects receive the levels of the IV in a different order. Latin square design = administer IV so it appears the same number of times in each position.

Question 32

Between-Groups (between-subjects) Designs:

Answer 32

different levels of the IV to different groups the compare on DV. Can expand with more than 2 levels of single IV or 2 or more IVs. When 2 or more IV = factorial design which allows more thorough information: can analyze main effects of each IV and interaction between IVs. Distinguish between main and interaction effects. Main effect = effect of 1 IV on the DV, disregarding effects of all other IVs.

Question 33

Interaction

Answer 33

= effects of 2 or more IVs considered together. Interaction occurs when the effects of an IV differ at different levels of another IV.

Question 34

Within-Subjects (repeated measures) Designs:

Answer 34

All levels of the IV are administered sequentially to all subjects. Comparisons of different levels of IV are made within subjects rather between subjects. 2 or more levels of IV, or 2 or more IVs. Single Group Time series design = effects of IV are evaluated by measuring the DV several times at regular intervals both before and after the IV is applied. Subjects are their own no treatment controls. Shortcoming = internal validity can be threatened by history. Helps control maturation since this can be detected in overall pattern of pre and posttest scores. Another type: 2 or more levels of IV are applied sequentially to each subject and the DV is measured after each level is applied. Problem is carryover effects (multiple treatment interference). ie: high dose may be effective because if followed low dose. Counterbalancing can control carryover effects, give different doses in different order. Disadvantage of within designs is that analysis of the data can be confounded by autocorrelation = performance on the post tests are likely to correlate with performance on the pretests. Can inflate the value of the inferential statistic (t or F) which results in increased probability of Type I error. Need special statistical techniques to analyze data with this design.

Question 35

Mixed Designs

Answer 35

combines between groups and within subjects methodologies. Common when measuring DV over time or across trials. Time or trials is an additional IV and is considered a within subjects variable because comparisons on the DV will be made within subjects across time or across trials. Compare effects of 4 levels of therapy by assigning patients to 1 of the levels and measure short and long term effects of therapy by administering assessment tool at 2 month intervals for 24 months.

Question 36

Single-Subject Designs

Answer 36

From behavioral psychologists; to investigate effects of an IV on behavior of 1 subject or small number of subjects or groups. What distinguish from groups designs: 1) each single subject design includes at least 1 baseline (no treatment) phase and 1 treatment phase. Each subject acts as his own no treatment control., 2) the DV is measured repeatedly at regular intervals throughout the baseline and treatment phases. Repeated measurement of the DV helps control any maturational effects.

Question 37

AB Design:

Answer 37

Single subject design. Most common and simplest. Single baseline (A) phase and single treatment (B) phase.

Question 38

Reversal Designs (ABA, ABAB, Etc.):

Answer 38

Single subject design. Expand the AB design to include more than 1 baseline or more than 1 baseline and more than 1 treatment phase.Because any expansion requires the withdrawal of the treatment during the 2nd and subsequent baseline phases, the extensions of the AB design are called reversal (withdrawal) designs. Advantage is that they provide additional control over potential threats to internal validity. If status on DV returns to initial baseline level during the 2nd can be more certain that any observed change in the DV is due to IV rather than to history. Inappropriate when withdrawal of a treatment would be unethical. Not conclusive if effects of IV persist rather than reverse when it is withdrawn as can’t be certain if observed effect on DV is due to IV or other factors. A phase, and then to its previous treatment levels during the 2nd

Question 39

Multiple Baseline Designs:

Answer 39

Use if reversal is inappropriate ethically or practically. Does not require withdrawal but involves sequentially applying the treatment to different behaviors of the same subject (multiple baseline across behaviors); to the same subject in different settings ( multiple baseline across settings); or to the same behavior of different subjects (multiple baseline across subjects). Once the treatment has been applied to a “baseline: it is not withdrawn from that baseline during the course of the study. Example: effects of self instruction on attention span with multiple baseline across settings. Train in self instruction and use when working on math in 3 different settings, measure the DV before and after at regular intervals. To be effective, the setting, behaviors or subjects chosen must be relatively independent.

Question 40

Descriptive Statistics

Answer 40

= used to describe and summarize data collected on a variable or the relationship between variables. ie. Use central tendency or scatter plot.

Question 41

Inferential statistics

Answer 41

are used to determine if obtained sample values can be generalized to the population from which the sample was drawn.

Question 42

-Continuous variable

Answer 42

can take on an infinite number of values on the measurement scale. ie: time

Question 43

-Discrete variable

Answer 43

assumes only a finite number of values. ie: DSM-IV diagnosis Another method distinguishes between 4 scales of measurement. Each scale differs in terms of the kind of info they provide and the mathematical operations they permit.

Question 44

Nominal Scale

Answer 44

= divides variables into unordered categories. Sex, religion, political affiliation. If number assigned it only acts as label, no order. Limitation is only math operation performed is to count frequency of cases in each category. Do not assume frequency implies a nominal variable. A variable is nominal when divided into categories and frequency in each category is compared. Frequency of aggressive acts and number of hours studied are ratio, not nominal data.

Question 45

Ordinal Scale

Answer 45

= divides observations into categories and provides info on order. Can say one has more or less or is midway. Ranks or Likert scales are examples. ie: strongly agree, agree, etc. with each being assigned a number. Limitation: can not say 10 is twice as much as 5.

Question 46

Interval Scale

Answer 46

= Order and equal intervals between successive points on the measurement scale. Scores on standardized IQ tests: can say difference between 90 – 95 is same as 100 – 105. Can do addition and subtraction. Can calculate a mean or standard deviation. Temperature on Fahrenheit of Celsius scale. Sometimes have zero point but it is arbitrary and not an absolute zero meaning absolute lack of or absence of the characteristic measured.

Question 47

Ratio Scale

Answer 47

= Most complex. Has order and equal intervals and absolute zero. 0 = complete absence. Makes it possible to multiply and divide. Can say one is 3 times more than another. and determine more precisely how much more of less there is Temperature on Kelvin scale, number of calories, # of correct items.

Question 48

Measures of Central Tendency

Answer 48

Usually want to describe data in single number which should convey maximum amount of information, summarize the entire set, and be a “typical” measure of all of the observations. Measures of central tendency most commonly used. 1. Mode (Mo) = score or category that occurs most frequently. Can be bimodal, or no unique mode. Easy to ID. Disadvantage is very susceptible to sampling fluctuations. Different sample modes can vary considerable. Mode is not useful for other statistical purposes and serves only as a descriptive technique. 2. Median (Md): Divides distribution in half. If distribution has an even number of numbers, it is the value that is between the 2 middle scores. Advantage, is if 1 score is extremely high or low, the value is not affected. Insensitive to outliers. Disadvantage is other quantitative procedures is limited and primarily descriptive. 3. Arithmetic Mean (M or X) = arithmetic average calculated by : M = X/N. M=mean, =add up, S=raw score, N = number of observations. Is least susceptible to sampling fluctuations. Is an unbiased estimate of the population mean. Can be used in a number of statistical procedures. Disadvantage is affected by magnitude of every score so if skew, mean is misleading.

Question 49

Choosing a Measure of Central Tendency.

Answer 49

Depends on data’s scale of measurement. Scale of Measurement Measure of Central Tendency Nominal: Mode Ordinal: Mode or Median Interval: Mode, Median, or Mean Ratio: Mode, Median, or Mean Generally use the measure of central tendency that lends itself to the greatest number of mathematical operations. But median is often used if skewed or missing data, especially at the high or low end of distribution, as it will be more representative. In a normal distribution , the 3 measures of central tendency are equal to the same value. In a positive skewed distribution, mean is greater, then the median and then the mode. In negative skewed = reverse.

Question 50

Measure of variability

Answer 50

indicate amount of heterogeneity or dispersion within a set of scores and includes the range the variance and the standard deviation.

Question 51

Range:

Answer 51

= Simplest. Calculated by subtracting the lowest score in the distribution from the highest score. Can be misleading if atypically high or low score.

Question 52

Variance (Mean Square):

Answer 52

Variance is more thorough measure of variability than range because its calculation includes all of the scores in the distribution rather than just the highest and lowest. Formula: Numerator = sum of squares or sum of the squared deviation scores. Sum of squares = subtracting the mean from each score to obtain deviation scores, squaring each deviation score and then summing the squared deviations scores. The size is affected by the amount of variability in a distribution and the number of scores. The more scores, the larger the sum of squares. To be useful the sum of squares is divided by N – 1. The result is the variance (mean square) or the mean of the squared deviation scores. Provides measure of the average amount off variability in a distribution by indicating the degree to which scores are dispersed around the distributions mean. The denominator for the variance is N when the variance for the population is being calculated; when a sample variance is being calculated, especially when it is going to be used an estimate of the population variance, the denominator is N – 1 because the sample variance tends to underestimate the population variance, and subtracting 1 from N reduces this bias.

Question 53

Standard Deviation:

Answer 53

more thorough measure of variability than range because its calculation includes all of the scores in the distribution Is a measure of variability that is the same unit of measurement as the original scores, so it is easier to interpret. Is the square root of the variance. SD can be interpreted directly as a measure of variability. The larger the SD, the greater the dispersion of scores around the distributions mean. Useful when comparing variability of 2 or more distributions. Can also be interpreted in terms of the normal distribution.

Question 54

Standard deviation, normal distribution and percentages w/in: 1SD, 2SD 3SD

Answer 54

If the shape is normal can draw conclusions about number of cases which will fall within limits defined by the standard deviation: 68.26% of scores fall between + 1 SD from the mean; 95.44% between + 2 SD from the mean; 99.72% between + 3 SD from the mean. If mean of 50 and SD of 5, and normally distributed, 68% of people have scores between 45 and 55. Also 84% have scores below 55. Can also select people with scores in top 16% = scores above 55.

Question 55

Inferential Statistics

Answer 55

Inferential statistics are used to make inferences about a population based on data collected from a sample drawn from the population and do so with a pre-defined degree of confidence. Can make inferences about relationships between variables in a population based on relationships observed in a sample through a sampling distribution.

Question 56

Standard error of the mean.

Answer 56

Which provides an estimate of the extent to which the mean of any one sample randomly drawn can be expected to vary from the population mean as the result of sampling error. = measure of variability due to effects of random error. Size of standard error of the mean is affected by the population SD and the sample size (N). The larger the population SD the smaller the sample size = the larger the standard error and vice versa. The smaller the sample size, the greater the probability ofor error when using a sample statistic to estimate a population parameter. For any given population, there is a “family” of sampling distributions with a different distribution for each sample size. A sampling distribution can be derived for any sample statistic, such as SD, proportions, correlation coefficients, difference between means, etc.

Question 57

The Logic of Hypothesis Testing

Answer 57

An inferential statistic would compare the obtained sample value to the appropriate sampling distribution and the results of the test would indicate whether an observed effect of the self control procedure was due to sampling error or to the effects of the procedure. Testing a research hypothesis about the effects of an IV on a DV involves the following steps: 1) Translate the verbal research hypothesis about the relationship between IV and DVs into 2 competing statistical hypotheses: the Null Hypothesis and the Alternative Hypothesis. 2) Conduct the study and analyze the obtained data with an inferential statistical test. 3) Decide on the basis of the results of the statistical test whether to retain or reject the statistical hypotheses.

Question 58

Hypothesis testing graph

Answer 58

Question 59

Type I error

Answer 59

= when investigator rejects a true null hypothesis. Thinks there is a difference, but really there is not. Probability of making a Type I error is equal to alpha. Increasing alpha from .01 to .05 increases probability of making a Type I error from 1 chance in 100 to 1 chance in 20, so investigator , setting alpha, has control over the probability of making a Type I error. It may also be increased by sample size being too small or when observations are dependent.

Question 60

Type II Error

Answer 60

when investigator retains a false null hypothesis. This is equal to beta (β ). Is not set by an investigator and cannot be directly calculated for a particular study, but it is influenced by : when alpha is low, when the sample size is small and when the IV is not administered in sufficient intensity. There is an inverse relationship between the Type I and Type II errors. As probability of making a Type I error increases, the Probability of making a Type II error decreases and vice versa.

If it is problematic to get a Type I error(rejecting true Null Hypothesis), choose level of significance that minimizes that (.01 instead of .05). In situation whereType II error should be avoided (retain a false null hypothesis), a larger level of significance is preferred = .1 or .05rather than .01.

Question 61

To maximize statistical power by :

Answer 61

• Increasing alpha, .05 instead of .01
• Increasing sample size. Correct decision more likely with 50 than 25 sample size.
• Increasing the effect size: Maximize the intensity of the IV or duration.
• Minimizing error: control systematic and random error by making sure the DV measure is reliable, reducing with-in group variability by controlling extraneous variables or by using a with-subjects design.
• Using a 1-tailed test when appropriate: it is more powerful when appropriate.
• Using a parametric test: such as t-test or ANOVA which are more powerful than nonparametric tests. Note that power is not the same as “confidence”. Power = ability to reject a false null and is affected by the size ofalpha. Power increases as alpha increases and vice versa. Investigator is concerned about power before the decision abut the null is made. Confidence = the certainty a researcher has bout the decision already made about rejecting the null. More confidence that decision was correct to reject null when alpha is small (.01 rather than .05).

Question 62

Inferential statistical test are classified as what two types?

Answer 62

parametric or nonparametric.

Question 63

Parametric test

Answer 63

include the t-test and ANOVA, are used to evaluate hypotheses about population means, variance or other parameters. Appropriate when variable of interest is measured on interval or ratio scale and when certain assumptions about the population distribution(s) are met. (1) value of interest is normally distributed in the population, (2) when a study includes more than 1 group, there is homoscedasticity: variances of the populations that the different groups represent are equal. Violation of these assumptions, (especially homo) can increase the probability of making a Type I or Type II error. Parametric tests are robust with regard to violation of their assumptions, some deviation from a normal curve of from homo will not necessarily invalidate the tests results. Most effective way to maximize the robustness is to have equal number of subjects in each group. Also increase by having a larger sample size and setting alpha at a lower level, .01 rather than .05. Parametric test are not as robust with regard to the assumption of independence of observations. Even a small amount of dependence among observations can increase the probability of Type I error above the probabilityindicated by alpha.

Question 64

Nonparametric tests:

Answer 64

Used to analyze data measured on nominal or ordinal scale. Do not make assumptions about shape of population distribution(s), considered distribution free. Used to evaluate hypotheses about the shape of a distribution rather than the distributions mean, variance or other parameter. Chi-square, Mann-Whitney U test and Wilcoxon matched-pairs test. Shortcoming: use less precise nominal or ordinal data and less powerful (less likely to reject a false null hypothesis with nonparametric test than with a parametric one). Always use parametric if can.

Question 65

critical value,

Answer 65

which is the number that corresponds to the boundary that divides the sampling distribution into rejection and retention regions. The magnitude of the critical value is determined by 2 factors: alpha and the degrees of freedom.

Question 66

Degrees of freedom (df)

Answer 66

determine the distribution’s exact shape. Degrees of freedom are the number of values or categories in a distribution that are “free to vary” given that certain values or categories are known or fixed. If there are 8 test scores and the sample mean is being used to estimate the population mean and is known, then 7 is degrees of freedom, since 1 score can not vary. Method to calculate df depends on the statistical test.

T-test for single sample, the sampling distribution is based on the sample size and df is derived from the total number of subjects (N-1).

When single sample chi-square test, different kind of sampling distribution is used. Based on the number of categories (level) of the variable and the df = derived from the total number of categories (C – 1).

Question 67

chi-square test

Answer 67

is used to analyze the frequency of observations in each category (level) of a nominal variable (or other variable that is being treated as a nominal variable). ie: number of people who prefer 1 of 4 candidates.

Used to determine if the distribution of observed (sample) frequencies is equivalent to the distribution of expected frequencies.The expected frequencies are predicted by the null hypotheses and reflect no difference between categories.

Use of chi-square requires the data to be reported in terms of frequencies, that expected frequency for any one category be no less than 5 and that observations be independent (subject can appear in only 1 category).

(For EPPP) Remember 2 forms of chi-square. The single-sample (single variable) and the multiple-sample (multiple-
variable). When counting the number of variables it is total of both DV and IV.

Question 68

Single-Sample Chi-Square Test:

Answer 68

= Goodness of fit test. Is used when descriptive study includes only 1 variable and data analyzed are number of observations in each category of that variable.

Summary: Single-Sample Chi-Square TestUse: One variable; nominal (frequency) data

2

Statistic: X

Df: (C – 1), where C = number of “columns”

(levels of the variable)

example: How many of 120 patients have 1 parent with Schiz, 2 parents with Schiz or no parent with Schiz. Df = (C–1)= (3-1) = 2.

Question 69

Multiple-Sample Chi-Square Test:

Answer 69

(“chi-square test for contingency tables”) is used when a descriptive or experimental study includes 2 or more variables and the data to be analyzed are the number of observations in each category.

Summary: Multiple-Sample Chi-Square Test

Use: Two or more variables; nominal (frequency) data

Statistic: X

Df: (C-1)(R-1), where C = number of “columns” and R = number of “rows”

Example: Also divide patients in groups according to one of 5 diagnosis. 3 categories for parents and 5 categories for patients. Is either a descriptive study with 2 variables or an experimental study with 1 IV (family background) and 1 DV (diagnosis). IV and DV are listed outside the table and inside the table is observed frequencies. This is different from interval or ration scale data and a t-test of ANOVA in which the IVs are listed outside the table and the number in the cells are the group means of the DV. (For EPPP)Be able to conceptualize this.

Question 70

Tests of Ordinal Data

Answer 70

Statistical test for ordinal include the sign test, the median test, the Kolmogorov test, the Mass-Whitney U test, the Wilcoxon matched-pairs test, the Kruskal-Wallis test. Last 3 are used to anlayze rank-ordered data.

Main thing to remember about the tests for ordinal data is they can be described as “nonparametric alternatives” to specific parametric tests.

• Mann-Whitney is nonpara alternative to the t-test for independent samples.
• Wilcoxon is nonpara alt to t-test for correlated samples, a
• Kruskal-Wallis is the nonpara alt to the 1-way ANOVA.

Question 71

Mann-Whitney U Test

Answer 71

Question 72

Wilcoxon Matched-Pares Signed-Ranks Test:

Answer 72

When study includes 2 correlated (matched) groups and the differences between the DV scores of subjects who have been in pairs are converted to ranks.

Summary: Wilcoxon Matched-Pairs Signed-Ranks Test

Use: 1 IV; 2 correlated groups

1 DV; rank ordered data

Statistic: U

Example: Assign patients to groups by matching them on basis of their scores on a measure of premorbid adjustment

and randomly assign members of each matched pair to either a drug x group or a placebo group. Because correlated

groups and IQ scores are not normally distributed, researcher calculates the difference score for each pair of matched

patients, converts the difference scores to ranks and uses the Wilcoxon test to analyze the data.

Question 73

Kruskal-Wallis Test:

Answer 73

Similar to Mann-Whitney U but can be used when a study includes 2 or more independent groups and the data is in ranks.

Summary: Kruskal-Wallis Test

Use: 1 IV; 2 or more independent groups

1 DV; rank-ordered data

Statistic: H

Example: Expands study by including 3 grups, 1 that will receive a high dose of drug X, 1 with low dose and 1 placebo.

If the IQ scores violate 1 or both assumptions for a parametric test, the researcher can convert the IQ scores to ranks and

use the Kruskal-Wallis Test to analyze the data.

Question 74

Tests for Interval and Ratio Data:

Answer 74

Student’s T-Test and the ANOVA are the most commonly used inferential statistical tests for variables measured on an interval or ratio scale. Several versions of both tests that are each appropriate for different research designs.

Question 75

Students T test for a Single Sample:

Answer 75

when only 1 group and the group (sample) mean will be compared to a known population mean. The population is acting as a no-treatment control group.

Example: self control increases achievement by training a random sample of 20 6th a test of academic achievement. Use t-test for a single sample to compare the mean achievement test score to the mean score for the population. df = 19.

Question 76

Student’s t-test:

Answer 76

Used to evaluate hypotheses about the differences between 2 means. 3 forms. Selection of a t-
test is determined by how the 2 means were obtained (sample and population, independent groups, or correlated groups) . A t-test can be used to analyze data collected from study involving more than 2 means, but it would be necessary to conduct more than 1 test. Problematic because the larger the number of statistical comparisons in 1 study, the more likely that a Type I error will be made = The greater the number of “experiments” the greater the experimentwise error rate. For this reason the ANOVA is preferred when more than 2 means are to be compared, ie when the IV has more than 2 levels.

Question 77

Student’s t-Test for Correlated Samples:

Answer 77

When 2 means to be compared have come from correlated groups. example with a study using within-subjects design in which a single group of subjects will be compared to itself before and after the IV has been applied. Or when subjects have been matched on an extraneous variable and members of each matched pair have been assigned to a different group or when subjects come to the study already matched as with twins. 20 ADHD children, tests them, then trains them, then retests them. df = 19

Question 78

Student’s t-test for Independent Samples:

Answer 78

20 ADHD students and randomly assigns 10 to treatment and 10 to control (no treatment). Train 1 group then test all. df = (N – 2) = 20 – 2 = 18

Question 79

Analysis of Variance:

Answer 79

ANOVA is used to compare 2 or more means. Advantage is that it makes all comparisons

of group means while holding the probability of making a Type I error at the level of level of significance set by the

investigator = ANOVA helps control the experimentwise error rate. Several versions. Most common is 1-way ANOVA

with 1 IV and factorial ANOVA with more than 1 IV.