Psych 225 Lecture 7Outline of Last Lecture 1) Stroop Results2) F distribution3) ANOVA4) T-test and descriptivesOutline of Current Lecture 1) Exam prep and questions2) Contrast t and LSD3) HSD4) A priori vs. Post Hoc tests Current LectureClass will get easier!Exam prep:-Berger demonstration chi square practice -utilize the course packet ex. pg. 371 (similar process as Stroop example) ex. Mood by attractiveness of target example pg. 48- won’t be covered in class ex. writing exercise on pg. 54-55 ex. pg. 56 series of questions to check your understanding: all questions are meant for no calculation-just look and analyze what’s given in the outputex. p. 38-39 line graph (describe effect) ex. pg. 37 a priori vs. post hoc ex. pg. 34-35 applying power principles: don’t need output, just think about QuestionsThese notes represent a detailed interpretation of the professor’s lecture. GradeBuddy is best used as a supplement to your own notes, not as a substitute.-What is the design of the next study?We are comparing valence (negative, neutral, positive) and association (weak or strong-semantically) All participants will see some negative, some neutral, and some positive word pairs Memory/meta cognitive study P’s will do JOL’s (judgments of learning) do you think you will remember cue word?Cue words and a target word -look at free recall & cued-look at source memory-sometimes the target word will be red and sometimes it will be blue: so we will ask participantsto identify the color of the target word -Run as a between subjects (the association) and a within (valence) thus it is an overall mixed design *get all the materials in lab next week*-Collect our data (3 participants, 6 with partner) and submit by 8pm on October 6th Class:-looked at type of presentation: consistent, inconsistent and control ; did not find support for 1sthypothesis but did for the second -looked at Course packet chart to determine a rough estimate of the p values for both the first and second experiment Pg. 15 in CP-determine the line to use by looking at degrees of freedom-had 5 participants per level -find degrees of freedom by: ~Mean Square Within df: (5-1) + (5-1) + (5-1)= 12 degrees of freedom (do N-1 per condition level) -df of contrast t test are based on df of MS within (12 df in f test and t test) -compared critical value to p when looking at degree of freedomContrast t test-a priori is before you collect data; know there are specific hypotheses you are going to test and that is what study was designed to do -post hoc is after you collect data ; exhaustive approachcontrast t test is an a priori approach-in think in terms of pair-wise comparisons -looking at least significant difference (LSD) is an algebraic equivalent to the t test (contrast t) -ethical approach: not pretend you made additional hypotheses when you didn’t: should only look at the least significant difference between means -solving for how small can our difference between means be and still be significant LSDConsider:1) what criterion of significance are we using 2) degrees of freedom (in this example=12) when looking at p 0.05, df=12 the critical value is 2.179 (so fill in the equation) thus the smallestmean we can have is 7.33 (and still be significant) -if we used p value of 0.01 (more restrictive) then t goes up (is bigger): have to show bigger differences /LSD for it to be significant -to make it smaller for it to be significant you would have to increase the degrees of freedom (bigger sample size, t shrinks, answer shrinks) Support our hypothesis?-with an LSD of 7.33:For Hypothesis 1: we do not have support with 2.8 (does not exceed 7.33) observed mean difference does not exceed LSDFor Hypothesis 2: we do have support with 7.8 (exceeds 7.33) observed mean difference exceeds LSDIf we got 7.33 exactly, we still have support (it can be equal or exceed to have support) Pg. 19: CP-consistent vs. control: .421 (chance we will get this by chance alone=.421 very very likely) ~contrast t p value on pg. 20=.421 -looking at inconsistent vs. control: .039 (knew it was less than .05)~contrast t p value on pg. 20=.04HSD=Honestly Significance Difference -need to consider an additional element (that we ignored in LSD-something that q needs that t didn’t): number of means -see pg. 16 in Cp-as means increase, the q level increases: the answer of HSD will increase*built in control over type 1 error rate (over experiments) (helps to tighten the control on type 1error when you do multiple tests) -ex. for this study: q=3.77 (when look at p of .05) and plugging this into the equation we get HSD=8.96-MSW for both HSD and LSD is 28.27Comparing all the means to all the means because it is a post hoc approach that is exhaustive -so we look at 10.6, 2.8, and 7.8(original means: 14.8, 25.40, and 17.60)-in the tukey HSD: p is bigger (.091)~as mean differences get larger, the p values get smaller -how do you guess what mean difference must be before it is significant?~answer: mean difference has to be between 7.8 and 10.6 (course packet pg. 19) Course Packet: pg. 37SEE COURSE PACKET*Purposea priori: confirm or falsify hypothesisPost hoc: explore, learn something *F testA priori: significance not necessaryPost hoc: p < or = to .05 (otherwise why continue to explore?)*p (Type II error) or beta error: we miss something that was actually therea priori: lowerpost hoc: (higher)*# of mean differencesa priori: fewer than possible; those that are theoretically groundedpost hoc: all possible #= k(k-1) divided by 2*Control of experimentiwise error ratea priori: literaturepost hoc: mathematical*Examplesa priori: contrast t, LSDpost hoc: Tukey HSD*Tablea priori: tpost hoc: q*magnitude of mean difference necessarya priori: smaller*powera priori: higher*p (mean difference r one more extreme) a priori: smaller Contrast t and LSD are the same mathematically (in pairwise comparisons)Rules for Contrast Coefficients1) assign a contrast coefficient for each level of the IV (each M) ~contrast t more versatile -if coefficient=0, that/those Ms drop from analysis 2) the sum of the coefficients must=03) sum of the positive coefficients must be 1.0 and the sum of the negative coefficients must be -1.0Impact of intervention of depression -H: after 2 months, P’s who received actual treatment will be less depressed than those who received no treatment : not pairwiselook at CP pg.
View Full Document