Study Guides (248,396)
Canada (121,508)
Psychology (1,711)

psych2840 Review Lectures 7-11.docx

20 Pages
Unlock Document

Psychology 2800E
Anthony Skelton

Psych2840: Lecture 7/8 Statistical tests for expreriments • What is a P-value: probability of data actually happening • Choice depends on o Research questions (compare/relate)\ o Desing (within/between/mixed) T-Test vs ANOVA • Both inferential statistics for research questions about group differences • Both provide a p-value o Can only use T-Test for 1-2groups/conditions o Can useANOVAfopr 2 or more groups • Can calculated by hand or using software Decision • How do you decide whether to reject or retain the null hypothesis? • Calculate p-value using appropriate test (ie T-Test,ANOVA, FactorialANOVA) o Learn to do this in stats class, you just need to know which test to use and which decision to make based on p-value • Compared p-value to alpha(how to make a decision on the hypothesis) o If P< alpha then reject the null hypothesis • Makes things statistically significant o If P>alpha then fail to reject null hypothesis • WONT BE EXPECTED TO CALCULATE P-VALUE BUT WILL BE EXPECTED TO MAKEADECISION Type I and Type II errors • Two ways to make a wrong conclusion/decision • Type I error: occurs when we reject the null hypothesis, but the null hypothesis is true o Like a false positive on a medical test o Detrimental because ppl take action on a mistaken finding • Type II error: occurs when FUCKKK Power • Probability of correctly rejecting the null hypothesis when it's false • Ways to increase the power of an experiment: o Increase the intensity of the experimental procedure o Use a less diverse population(or within design) o Use more precise measurement o Use larger sample size WARNINGS •Proved(instead use supported) •Insignificant (instead use non-significant) •Accept the null hypothesis (instead use fail to reject) •Accept the research hypothesis (instead use reject the null hypothesis) •Will take marks of exam; DON’T USE TERMS Parametric Tests •Require assumptions about the population •Assumptions - criteria that are met, ideally before conducting a hypothesis test •Robust hypothesis test- one that produces valid results even when all assumptions aren't met Parametric statistics Assumption •Participants are randomly selected •Pop' distribution is approximately normal •DV is a scale measurement Breaking theAssumptions •OK if we are cautious about generalizing •OK if sample includes at least 20 scores (within design) or 40 for between subjects •Usually OK if the data are not clearly nominal or ordinal Non-perametric statistics •Statistical analyses that are not based on a set of assumptions about the pop' • Use a non-parametric test when you can't use a parametric test(no choice) •When you use a non-parametric stats: o When DV is nominal o When DV is clearly ordinal o When sample size is small (<20) and we suspect the distribution of pop' is skewed •Follow same steps of hypothesis testing •Non-parametric stats expand the range of variables available for statistical analyses (pro) Limitations • Typically have less statistical power (increased risk of type II error) •Effect size measures are not typically available LOOKAT PICTURE OF TABLE ON PHONE Ethics: Basic Principles •People generally shouldn't be involved in research o w/o informed consent o w/ increased risk of harm •Principles o Respect for personsAUTONOMY o BENEFICIENCE (do good to) o JUSTICE Deceit • Milgram's experiment: studying how people obey authority even if it conflicts w/morals (shocking a "confederate" because the researcher told them to) • If deception is used: include a statement that "the research cannot be fully described at this time, but at the conclusion of participation, a explanation will be provided" • Be mindful of belief perseverance- the debriefing may not be effective in undoing the effects of deception o i.e. suicide notes Observational vs. Survey Research • Is the variable better measured as a self-report or observation? Propose an exact op def. • Apolitical psychologist wants to measure the percentage of ppl who vote in elections in rural Ontario(self-report) • Acognitive psychologist wants to measure how often people send texts while driving in London (observational) • Asocial psychologist wants to measure how much jealousy women feel in relationships w/ their partners (self-report) Choosing Question Formats • Open-ended questions • Forced-choice format • Likert scale • Semantic differential format (same as Likert scale; use other terms rather than agree, disagree, strongly agree/disagree) Writing well-worded questions • Leading questions • Double-barreled questions: two questions in one(poor construct validity; not sure what single questions is being asked) • Double negatives (vs negatively worded items): "I don’t not wanna be here"(poor construct validity; not sure what single questions is being asked) • Question Order: earlier questions affect how ppl answer the later questions Lecture 7: Research Methods in Psychology   Statistical Validity   ­ Statistical significance unlikely that results you found just occurred by chance. Doesn’t  mean there’s meaningful difference.  • What is the probability that the results occurred by chance? • Determine using p­value   • Strength • How strong is the effect/association? • Determine using effect size?   ­ Power • What is the probability of finding an effect in the experiment if there really is an  effect? • Determine using power    Why Statistics?   • Descriptive: summarize data so that we can understand and use it.  • Inferential­ are the results generalizable? How likely is it that the results  happened by chance? (no matter which taste you do it will give you a p­value, p­value  will tell you if things just happened by chance)   Types of Questions • Comparing: Is there a difference between groups? • Relating: Is there a relationship between variables?   Central Tendency • One of most important ways to understand a data set­ the “typical” score • We can use the mean, median, or mode • Choose one that best represents your data!\   Mean • most commonly reported measure of central tendency   Median • 2  most popular measure of central tendency   Mode • most common score of all scores in sample   How Outliers effect measures of central tendency • Outlier­ an extreme score that is either very high or very low in comparison with  rest of scores in sample • Mean is affected by outliers (non­resistant) • Median and mode are resistant to outliers   So which one to use? • Goal: want to choose most informative measure that is appropriate for data • Generally, mean is most informative, then median, then mode • But, choice affected by: scale of measurement and shape distribution  • For nominal data, always use mode • For ordinal data, can use median or mode depending on shape of distribution • For scale data, can use mean, median, or mode (depending on shape of  distribution). Interval or ratio counts as scale    Variability • Another important way to understand a data set – how spread out the scores are • The most commonly reported measures of variability are the range, standard  deviation and quartiles  • Again, choose one that best represents the data!    Range • Distance that a distribution of scores covers on number line   Standard Deviation • Typical amount that each score varies or deviates from mean • Always positive • Non­resistant to outliers  • Square root of variance  Quartiles • One of three points that divides a distribution into fourths • Q1: point that has 25% of scores below it • Q2: median • Q3: point that has 75% of scores below it So which to use? • Goal: again, want to choose the most informative measure that is appropriate for  data • Generally, standard deviation is most informative, then quartiles and range • Again, choice affected by: scale of measurement and shape of distribution    So which one to use? • Measures of variability pair with measures of central tendency, so… • If using mean, then use standard deviation • If using median, then use quartiles or rage • If using mode, alas, you are SOL (ponder why this is)    be able to tell which one is right to use    Inferential Statistics • Logical process of using data from sample (whose characteristics are known) to  make inferences about a population (whose characteristics are unknown)  Hypothesis testing • Assume there is no effect­ null hypothesis • Collect data • Calculate probability of data (or more extreme data) if null hypothesis is true • Decide whether to reject or retain the null hypothesis    Formulate hypothesis • Testable statement about a presumed relation between two or more variables • Falsifiable (can be disproven but not proven) • About populations (not samples)    2(lottery status: win, loss) Between Ss design. DV: happiness (1­100) Research hypothesis: Winning the lottery will make people happier 6 months later compared  to those who lost.    Comparing Hypotheses ­ Null hypothesis H (little o) • The IV doesn’t have an effect on the DV • There’s no relationship between the variables  ­ Research hypothesis H (little 1) (aka alternative hypothesis H little A) • The IV doesn’t have an effect on the DV • There’s relationship between the variables    • Formulate null and research hypotheses to set them up against one another  • Mutually exclusive • Exhaustive (all options have been covered) ­ Data lead us to conclude one of two things  1. Reject the null hypothesis (this is what you want to do,) or 2. Fail to reject the null hypothesis    Logic of hypothesis testing • Because the null and research hypotheses are mutually exclusive and exhaustive,  either the null hypothesis is true or the research hypothesis is true • We can’t directly show that the research hypothesis is true • However, if we can show that the null hypothesis is false we can eliminate it from  consideration and by default demonstrate that the research hypothesis must be true  • If null hypothesis is true then both populations are the same    What the heck is a p value? • P value *** is probability of the data given the null hypothesis is true p  (data/null) • Hypothesis testing works in same way • What if the null hypothesis is true­ what implications follow from that? • Not the same as probability of the null hypothesis given the data! – Which we do  not calculate    Statistical tests for experiments • All will provide p­value • Choice depends on: • Research question (compare/relate) • Design (within/between/mixed)  • Number of IVs • Number of levels  • Scale of measurement    Alpha and p - Alpha • Pre­defined cut off point (=.05, or sometimes .01) • Probability of committing a Type I error (false positive) • Decision point to reject or retain null hypothesis    - P-value • Calculated using the appropriate inferential statistic  • Probability of the data (or more extreme data) if the null hypothesis is ture • Statistical validity – Is result statistically significant?   Decision • How do you decide whether to reject or retain the null hypothesis? • Calculate p­value using appropriate test (e.g. T­test, anova, factorial anova): you  will learn to do this in stats class, you just need to know which test to use and which  decision to make based on p­value  • Compare p­value to alpha (.05) • If p  or equal to alpha then fail to reject null hypothesis      Handout 2 Research Methods Day 7 Statistical Validity: Statistical significance – unlikely results occurred by chance - what is the probability that the results occurred by chance? - determine using p-value Strength - how strong is the effect/association - determine using effect size Power - what is the probability of finding an effect in the experiment if there really is an effect? - determine using power - to determine power, we use power Why Statistics? Descriptive – summarize data so that we can understand and use it Inferential – are the results generalizable? How likely is it that the results happened by chance? Types of Questions: Comparing - is there a difference between groups/conditions Relating - is there a relationship between variables Use a T-Test For my assignment* Central Tendency: - one of the most important ways to understand data set “the typical score” - we can use the mean, median, or mode as an indicator of central tendency - choose the one that best represents the data! Know which one to use for the data for the exam* Mean: Average of Datascores Median: The middle score of all the scores in a sample when they are arranged in an ascending order Mode: Most common score How outliers effect measures of central tendency: Outlier – an extreme score that is either very high or very low in comparison with the rest of the scores in the sample - mean is affected by outliers (non-resistance) - median and mode are resistant to outliers So which to use? Goal – want to choose the most informative measure that is appropriate for the data - generally, the mean is the most informative, then median, then mode But: choice is affected by scale of measurement - scale of measurement - shape of distribution So which one to use? For nominal data, always use mode For ordinal data, can use median or mode (depending on the shape of the distribution) For scale data (interval or ratio), can use mean, median, or mode (depending on the shape of distribution) - if you have outliers, then mean is not a good option Variability: - another important way to understand a data set – how spread out the scores are - the most commonly reported measures of variability are the range, standard deviation, and quartiles - again, choose the one that best represents the data! Range: - the easiest measure of variability to ccalculate - the least informative (does not tell if scores are clustered or spread out between endpoints) Definition: the distance that a distribution of scores covers on a plot Standard Deviation: Definition – the typical amount that each score varies (or deviates) from the mean - standard deviation is in the original unit of measurement (Same as mean) - standard deviation and variance are always positive values - non resistant to outliers Quartiles: Definition: one of the three points that divides a distribution into fourths Q1 – point that has 25% of scorers below it Q2 – median – the point that has 50% of scores below it Q3 – point that has 75% of scores below it So which one to use? Goal: again, want to choose the most informative measure that is appropriate for the data - generally, standard deviation is most informative, then quartiles, and range Again, choice is affected by: - scale of measurement - shape of distribution So which one to use? Measures of variability pair with measures of central tendency so... If using mean, then use standard deviation If using median, then use quartiles or range If using mode, alas, you are SOL Inferential Statistics: - logical process of using data from a sample (whose characteristics are known) to make inferences about a population (whose characteristics are unknown) Hypothesis testing: 1. Assume there is no effect – null hypothesis 2. Collect data 3. Calculate probability of data (or more extreme data) if the null hypothesis is true 4. decide whether to reject or retain the null hypothesis Formulate hypothesis: - testable statement about a presumed relation between two or more variables - falsifiable (can be disproven but not proven) - about populations (not samples) Comparing Hypothesis: Null hypothesis Ho - The IV does not have an effect on the DV - there is no relationship between the variables Research hypothesis H1 (aka alternative hypothesis Ha) - the IV does have an effect on the DV - there is a relationship between the variables Comparing hypotheses: Formulate null and research hypotheses to set them up against one another - mutually exclusive (do not relate to each other, no overlap) - exhaustive (no other option) Data leads us to conclude one of the two things 1. Reject the null hypothesis (this is what you want to do) or, 2. fail to reject the null hypothesis Note: never use the word accept in hypothesis testing “mu” is u = population mean Ho: u winners ≤ u losers Ha: u winners > u losers Logic of hypothesis testing: Because the null and research hypotheses are mutually exclusive and exhaustive, either the null hypothesis is true or the research hypothesis is true - we cannot directly show that the research hypothesis is true - however, if we can show that the null hypothesis is false, we can eliminate it from consideration and by default demonstrate that the research hypothesis must be true What the heck is a p value? Exam* Definition – the probability of the data given the null hypothesis is true p(data/null) - not the same as the probability of the null hypothesis given the data! – which we do not calculate Statistical tests for experiments: All will provide a p value Choice depends on - research question (compare/relate) - design (within/between/mixed) - number of IV’s - number of levels - scale of measurement Alpha and p: Alpha - pre-defined cut-off point (.05, or .01) - probability of commiting a Type 1 error (false positive) - decision point to reject or retain null hypothesis P-value - calculated using the appropriate infernal statistic - probability of the data (or more extreme data) if the null hypothesis is true - statistical validity – is the result statistically significant? Decision: How do you decide whether to reject or retain the null hypothesis? Compare p-value to alpha (.05) - if p < alpha then reject the null hypothesis - if p ≥ alpha then fail to reject the null hypothesis P = .15 > ∞ = .05 Means we fail to reject Ho  Lecture 9 T test VS ANOVA • Both give a P value • Only use a T test for one or two conditions • ANOVA with 2 or more conditions • Can calculate by hand or using statistical software Decision • How do you decide whether t reject or retain 
More Less

Related notes for Psychology 2800E

Log In


Join OneClass

Access over 10 million pages of study
documents for 1.3 million courses.

Sign up

Join to view


By registering, I agree to the Terms and Privacy Policies
Already have an account?
Just a few more details

So we can recommend you notes for your school.

Reset Password

Please enter below the email address you registered with and we will send you a link to reset your password.

Add your courses

Get notes from the top students in your class.