¡GoConqr estrena diseño! Cuéntanos qué te parece.

Regístrate gratis

Fichas

Lecture 9

Samar Castelli Inglés 93 fichas

Fichas del conjunto (93)

  1. 1What does single-factor mean in ANOVA? (also written as one-way)One treatment/independent variable
  2. 2What does between-subjects in ANOVA mean?There are independent groups; subjects in each treatment group are completely independent of the subjects in the other treatment groups. Each subject receives one treatment (is exposed to one level of independent variable) only.
  3. 3What if you wanted to run the experiment so that each subject received all three types of memory enhancement method (plus provided a control measure)?You would need to conduct a repeated measures or within-subjects design.
  4. 4What is a repeated measures or within subjects design?Where each subject receives ALL treatments (is exposed to all levels of the independent variable).
  5. 5What is the most common variable that adopts a repeated measures designTime
  6. 6What do we call a study, where the dependent variable is measured at several time points for each subject, often over a relatively long period of time.Longitudinal
  7. 7What do we call the IV in a classical experimental methodology where people are recorded at a baseline, then a treatment is implemented, then the participants are recorded again and possibly followed-up to establish if results were maintained?Time or phase
    This is a repeated measures design because the same people are measured a different time points. Usually, there is some independent variable (often called a within-subject factor) that changes with each measurement.
  8. 8What is the IV and the levels of the IV in the scenario?IV = time
    Levels: different time points that are measured
  9. 9Provide an example of a repeated measures research scenarioWhat is the effect of a blood-pressure drug over three time periods—morning, noon, and evening?
  10. 10What is the IV and DV in this scenario?o IV: Time (3 levels—morning, noon, evening)
    o DV: Blood pressure measurement
  11. 11What type of design is this, how would it be described?A single-factor (or one-way) within-subjects (or repeated measures) design. A design where each participant is measured/tested on repeated occasions. Each participant provides data for each level of the independent variable.
  12. 12Why is the design still a single-factor/one-way design?Because we still have only one independent variable: time.
  13. 13How would the data be setup in SPSS?Data set-up is exactly the same as for paired-samples t-test. No separate grouping variable—the levels of the IV are represented by three separate columns/variables in the datasheet.
  14. 14What are the SPSS steps to obtain a graph to describe the data?1. Graphs
    2. Legacy dialogues
    3. Bar
    4. Simple
    5. Summaries of separate variables (vs summaries for groups of cases for a independent groups design)
    6. Define
    7. Move variables from left to bars represent)
    8. Ok
  15. 15What is the null hypothesis in this scenario?morning = noon = evening
    (and any deviation that we see from this in the sample is just sampling error).
  16. 16What is the null hypothesis always about?1. What's going on in the population
    2. A statement of identity
  17. 17What's the alternate hypothesis in this scenario?At least one of the means is different.
    *Similar to between-subjects ANOVA*
  18. 18How do you run a repeated measures ANOVA in SPSS?1. Analyse
    2. General liner model
    3. Repeated measures
    4. Add IV name (time) into within-subject factor name
    5. Add number of levels (3)
    6. Add
    7. Define
    8. Move over variables to corresponding levels on the right hand dialogue box
    9. Ok
  19. 19What is the independence of error also called? Can't make this assumption in repeated measures!!The assumption of independence
    Can't make this assumption in repeated measures!!
  20. 20Provide a definition of the assumption of independenceEvery data point is separate from the other data points and data points don't influence each other
  21. 21Why is the assumption of independence of error relevant (an issue) in a repeated measures design?Because the same participants are being used, therefore earlier scores can impact later scores.
  22. 22Why is there such a large amount of output for a repeated measured design in SPSS?Because of the assumption of independence. SPSS tries to fix the problem associated with this assumption in multiple ways. Therefore, SPSS provides eight different ways of testing the same hypothesis.
  23. 23In what other test does SPSS do this?t-tests
    Provides the two versions HOV assumed/not assumed.
  24. 24What's an assumption associated with repeated measures?The assumption of Sphericity.
    Refers to the condition where the variances of the differences between all possible pairs of within-subject conditions (i.e., levels of the independent variable) are equal.
  25. 25What statistic is often used to answer the MAIN hypothesis in this research design?Wilks Lamba
  26. 26What ratio is provided in addition to the Wilks Lamba statistic?F ratio
    (a rudimentary version)
  27. 27How would the F ratio be written up?F(2, 6) = 21.23, p = .002.
    Remember to italicise!
  28. 28How would you write up the entire result?Add the Wilks Lamba statistic
    (L in symbols/looks like upside down V_
     = .12, F(2, 6) = 21.23, p = .002
  29. 29How is the Lamda statistics different from other statistics covered in the course?Gets more significant as it gets smaller.
  30. 30What are the APA conventions in terms of Lamda?Doesn't take a leading zero and is reported to two decimal places
  31. 31What does the following result allow us to conclude?  = .12, F(2, 6) = 21.23, p = .002Read like all other tests, based on significance.
    There is a significant difference somewhere, across the three times. * At least one of the means is significantly different.
  32. 32How would this be written up so far?A single-factor within-subjects ANOVA found a significant difference in the blood pressure scores across the three times of day,  = .12, F(2, 6) = 21.23, p = .002.
    Remember to italicise!
  33. 33How is the process similar to a between-subjects ANOVAFollows the same steps. Must obtain a significant result before conducting post-hoc tests to find the source of the difference
  34. 34Can you use the same post hoc tests that are used for a between-subjects design with repeated measures?No
  35. 35How do you carry out the alternative post hoc tests in SPSS?Same procedure as before but under options you:
    1. Under estimated marginal means you move the DV to Display means For
    2. Check compare main effects box
    3. Select Bonferoni in the drop down menu
    4. Continue
    5. Ok
  36. 36Where does the post hoc output start?After the phrase Estimated Marginal Means
  37. 37What does the first table display?Means
  38. 38Where are the results displayed?Under pairwise comparisons
  39. 39How is the output different to the between-subjects outputsLevels of IV are displayed numerically, not by name.
  40. 40What post hoc test do you select for repeated measures?Bonferoni
  41. 41What do we have to keep in mind in relation to post hoc test and ANOVA (repeated and otherwise)Because post hoc tests are run to confirm where the differences occurred between groups, they should only be run when you have a shown an overall statistically significant difference in group means
  42. 42How do you write up a repeated measures ANOVA?A single-factor within-subjects ANOVA found a significant difference in the blood pressure scores across the three times of day,  = .12, F(2, 6) = 21.23, p = .002. Post-hoc testing using paired samples t-tests with Bonferroni-adjusted  levels revealed a significant difference between blood pressure at morning and noon, p = .001, and between morning and evening, p = .015, but not between noon and evening, p = .23.
  43. 43What is the full name of the post hoc test?Paired samples t-tests with Bonferroni-adjusted alpha (use symbol) levels
  44. 44Why do we refer to the post hoc test in this way: Paired samples t-tests with Bonferroni-adjusted alpha (use symbol) levelsBecause we conducted three paired sample t-tests but manipulated the critical alpha level to take into account the use of a post hoc procedure.
  45. 45What needs to be reported with post hoc tests?Sig. level
    ONLY
  46. 461. What source of error is removed in a within-subjects design? & 2. What is the outcome of this?1. Individual differences between-groups. There can be no individual differences between groups when the same subjects are in each group.
    2. By removing a source of error, the effect of the treatment/IV is highlighted, and the test is more powerful.
  47. 47What are the advantages of a within-subjects design?1. Individual differences between-groups. There can be no individual differences between groups when the same subjects are in each group.
    2. Fewer subjects are required to obtain the same amount of data. For example, to get 40 scores across the four groups with a between-subjects design requires 40 subjects. To get the same amount of data with a within-subjects design requires only 10 subjects.
  48. 48What are the main problems with repeated measures designs?1. Drop-out: If you fail to obtain one score from any one subject, you have to remove all that subject’s data from the analysis.
    2. Practice/order/carry-over effects: Receiving one type of treatment or being exposed to one type of material might lead to an improved or varied performance in the other conditions, simply because of the initial exposure. What happens to you in an early stage of the study might affect how you perform in a later stage. Major problem.
  49. 49In which research scenario is this a particular problem?Drug trials.
    We don't know if the first drug is no longer impacting the participant. Thus, repeated measures are not used in drug trials routinely.
  50. 50Why is time such a popular repeated measures factor?it's not affected by carry over effects
  51. 51What are carry-over/practice/order effects, and how can they be dealt with?An effect that "carries over" from one experimental condition to another. Whenever participants perform in more than one condition (as they do in within-subject designs) there is a possibility of carryover effects.
    How to deal with this issue:
    Counterbalance or randomise the testing order:
    Make sure that different subjects receive the treatments in different orders. This guarantees that any order effect is at least randomised rather than systematic, but it does not remove the overall problem.
  52. 52What does counterbalancing aim to do?Balance out the carry over effects and allows the research to detect carry over effects
  53. 53Does counterbalancing prevent carry over effects?No
  54. 54What's the difference between counterbalancing and randomising?Changing the order in a very systematic way so that every possible order is represented. randomising means randomly choosing participants plus counterbalancing.
  55. 55If order effects make a within-subjects analysis unusable, what can you do?Use a matched samples design/randomised blocks design, which is an attempt to mimic a fully within-subjects design with different subjects.
  56. 56What is a matched samples or randomised blocks design ?With a randomized block design, the experimenter divides subjects into subgroups called blocks, such that the variability within blocks is less than the variability between blocks. Then, subjects within each block are randomly assigned to treatment conditions.
  57. 57What problem does it avoidMatches people on crucial variables to control sources of error whilst also avoiding carry over effects.
  58. 58Is a randomised block design common in cognitive psych?No
  59. 59What does matching in a randomised block design involve?Matching subjects on variables that we believe are going to have some relationship with the intervention/treatment/ technique.
    E.g. for a weight loss program:
    1. pre-program weight
    2. their body fat %
    3. their level of fitness
    4. their age
    5. their sex
  60. 60How do you match?Select x number of subjects who are as similar as possible on the matching variables. For example, four men, aged between 18 -25, between 20-25% overweight, who have an average level of fitness. Then randomly allocate these four men to the four experimental conditions.
  61. 61What are the advantages and problems associated with a randomised block design?Advantages:
    1. No carry-over/order/practice effects.
    2. Can analyse the same as a fully repeated measures design.
    Problems:
    1. Matching can be difficult.
    2. If you lose one subject in a block you generally lose the whole block.
  62. 62NON-PARAMETRIC ANOVANON-PARAMETRIC ANOVA
  63. 63What's the non-parametric equivalent of a single-factor between-subjects design?The Kruskal-Wallis test
  64. 64Under what circumstances would you use this test?DV is ordinal
  65. 65What's an example research question and statistical question in this scenario?o Research question: Are there differences among races in terms of how happy people report their lives to be?
    o Statistical question: Is there a significant difference among the three racial groups on the ranks associated with the scores on how happy people self-report their lives to be.
  66. 66What's the null in this scenario?H0: u(rank)white = u(rank)black = u(rank)other
    AND
    any deviation from this that we see in our sample is simply sampling error.
  67. 67How do you run the Kruskal-Wallis test?1. Analyse
    2. Non-parametric tests
    3. Legacy dialogues
    4. K independent samples
    5. DV goes into test variable list
    6. Grouping variable goes into grouping variable (but must enter highest and lowest race codes)
    7. Ok
  68. 68How would you write up the results of a Kruskal-Wallis non-parametric ANOVA?A single-factor between-subjects Kruskal-Wallis non-parametric ANOVA found a significant difference among the three race groups for the means associated with the ranks for self-reported happiness, x2 (2, N = 1504) = 18.37, p < .001.
    Italicise!
  69. 69Recap: what should a write up cover?1. Description of procedure used
    2. Summary of test result
    3. Statistical result after a comma
  70. 701. What does the Kruskal-Wallis test tell us? 2. Are post-hoc tests possible?1. That something is going on—that is, at least one group is different from the rest—but it doesn’t reveal the source of the difference.
    2. Yes, but must be done manually.
  71. 71How do you carry a manual test for a single-factor between-subjects non-parametric ANOVA1. Run every possible pairwise comparison
    2. Use a non-parametric equivalent of an independent samples t-test to run these tests (Mann-Whitney U test)
    3. Adjust the critical p value (i.e., the  level) manually.
    4. alpha = .05 divided by the number of post-hoc tests you run. In this case, 3.
    5. .05/3 = .017
  72. 72How do you run a Mann Whitney U test?1. Analyse
    2. Non-parametric tests
    3. Legacy dialogues
    4. 2 Independent samples
    5. Add variables
    6. Add groups that you want to compare under define groups
  73. 73Why do we need to adjust the alpha associated with the Mann Whitney u test?Because this is done automatically with “normal” post-hoc tests, such as Tukey, but we need to do it manually with non-parametrics) as they are not 'normal' post hoc procedures.
    We do this because post hoc tests are conservative due to the exploratory nature of post hoc tests.
  74. 74What is this adjustment called?This is called a Bonferroni correction or adjustment.
  75. 75How do you obtain the corrected significance levels?alpha = .05 divided by the number of post-hoc tests you run.
  76. 76How would you write up the results of a single-factor between-subjects non-parametric ANOVA post hoc test ?Post-hoc testing involved running all possible pairwise comparisons among the three groups using Mann-Whitney U tests with a Bonferroni adjusted  = .017. A significant difference at p < .001 was found between the white and black respondents. No significant differences were found between the white and other respondents, p = .08, or between the black and other respondents, p = .78.
  77. 77Are parametric or non-parametric tests more statistically powerful?Commonly help view that non-paramnateric tests are not as powerful as their parametric counterparts.
    However, there are contradictions in the literature.
  78. 78What's the non-Parametric equivalent of a single-factor Within-subjects ANOVAThe Friedman Test
  79. 79How is the data setup in a non-Parametric equivalent of a single-factor Within-subjects ANOVA?Same as paired sample t t-test, with the repeated measure each with their own column.
  80. 80Provide an example of a statistical questionIs there a statistically significant difference in how exciting, routine or dull people report their lives to be over the last three years.
  81. 81What would be the null?Null hypothesis:
    H0 u(rank)this year = u(rank) last year = u(rank) two years ago
    and any deviation from this that we see in our sample is simply sampling error.
  82. 82What would be the alternative hypothesis in this scenario?At least one of the groups has a mean rank that is different from the rest.
  83. 83How do you run the Friedman’s ANOVA?1. Analyse
    2. Non-parametric tests
    3. Legacy dialogues
    4. K Related samples
    5. Add all variables of interest
    6. Ok
  84. 84How would you write up the results?A single-factor within-subjects Friedman non-parametric ANOVA found a significant difference over the three times for the means associated with the ranks for self-reported excitement in life, x2 (2, N = 971) = 200.826, p < .001.
  85. 85What statistic is used to report the non-parametric ANOVAs?X2
  86. 86What are the steps involved in running post-hoc tests after a significant Friedman test:1. Run every possible pairwise comparison among the three groups.
    2. Use a non-parametric equivalent of a paried samples t-test to run these tests (Wilcoxon signed-rank sum test)
    3. Adjust the critical p value (i.e., the alpha level) manually.
    4. Alpha= .05 divided by the number of post-hoc tests you run. In this case, 3.
    5. .05/3 = .017
    6. Each Wilcoxon test must have an observed p < .017 in order to be significant. (This is done automatically with “normal” post-hoc tests, such as Tukey, but we need to do it manually with non-parametrics)
  87. 87What is the adjustment called?A Bonferroni correction.
  88. 88How do you run the post hoc test for a single-factor within-subjects ANOVA:1. Analyse
    2. Non-parametric tests
    3. Legacy dialogues
    4. Two related samples
    5. Add the pairwise comparisons
    6. OK
  89. 89What are the names of the post-hoc tests that are conducted for the two types of non-parametric ANOVAs that were discussed in class?Mann-Whitney U tests with a Bonferroni adjusted alpha (sign) =
    Wilcoxon Signed Ranks test with a Bonferroni adjusted alpha (sign) =
  90. 90Are the findings that you would reach from a parametric test similar to the findings you would obtain from a non-parametric test?Yes, the same
  91. 91Recap: when would you use an ANOVAWhen you want to compare more than two groups
  92. 92What would you need to divide the critical alpha by when conducting non-parametric equivalents of post hoc tests for 3 groups? 4 groups? 5 groups?3 groups? 3
    4 groups? 6
    5 groups? 10
  93. 93Are there non-paramentrcu equivalents of planned comparisons? Why?No, because non-parametrics are more limited in terms of design and the types of questions they can address.

Información

Descripción
Fichas sobre Lecture 9, creado por Samar Castelli el 06/08/2018.
Creado por
Samar Castelli · ago 2018
Actualizado
12 ago 2018
Visitas
2
Valoración

Más sobre Repeated Measures ANOVA

Escríbenos

¿Sobre qué es tu mensaje?

¿Qué te parece el nuevo diseño?

¿Te gusta?