Study Guides (248,519)
Canada (121,605)
Psychology (700)
PSY201H1 (26)
Quiz

Lecture 10 - The t Test for Two Independent Variables

4 Pages
101 Views
Unlock Document

Department
Psychology
Course
PSY201H1
Professor
Alison Luby
Semester
Fall

Description
PSY201H1F L10; Nov. 24, 2011 Assignment Info • Only need to go as far as text for assignment • Use APA style • Additional office hrs SS4040 5-7pm on Mon • Only paraphrasing last paragraph of intro and methods ○ When citing, put (assignment introduction) or (assignment methods) beside EACH piece of info (Glickstein, 2007) • When refer to study referenced in the given http://tvpclub.blogspot.com/2007/10/lies-damned-lies- intro/methods: and-statistics-part-4.html ○ Ex. Lustig, Haser, and Tonev (in press) say … • How to know if significantly different pops, overlap (assignment introduction). • Like all hypothesis tests, general purpose of indep- • No pg limit, keep as short as possible measures t test: ○ Clear, concise, accurate ○ Determine whether sample mean dif obtained • No need for a running head but do include a title in research study indicates real mean dif btwn page the two pops (or treatments) or whether • No abstract obtained dif is simply result of sampling error • Report findings from math with APA style, use text and APA manual for style Assumptions? • No outliers Ch. 10: The t Test for Two Independent ○ Trim? Leave out? Resample? ○ Important to have no outliers so data not Variables Comparison of Two Samples skewed • Ex. Is there a dif in amt of aggressive bhvr in girls • Homogeneity of variance – F-Max test vs boys at age 11? • Similar sample sizes • What is effect of 0 mg vs 10 mg of particular drug ○ Otherwise pooled variance (larger n has larger on reducing seizures in rats? effect) • Which is more effective, cognitive bhvr therapy • A testable preconceived hypothesis… (CBT) or drug therapy for treating depression in ○ If want to test something hadn’t originally young teenagers? planned for, can, but ideally the original data • Concerns mean difs btwn 2 separate & indiv grps should become pilot data & need to grab a new sample to test this hypothesis Independent-Measures Design • Aka. Between-Subjects Design Resampling techniques (don’t need to know) • Expands single-sample t test to test for difs btwn two separate and independent grps • Evaluate mean difference between two pops or btwn two treatment conditions • Unknown population means (µ) & standard deviations (σ) ○ Estimate from sample data • Might not know anything about parent pop distrs but if samples representative then we know more • Normally pop variance, standard deviation, mean unknown Notation • Use subscripts to specify sample st • 1 sample: n ,1M ,1SS 1 • Resampling – sampling w replacement via • Second sample: n ,2M ,2SS 2 computer • Population mean: μ a1d μ 2 • H0: μ1-μ2= 0 (no mean difference) ○ Do since can’t apply t test to non-normal distr • H1: μ1-μ2≠0 (mean difference) • When all you sampled is all you can get, Hypothesis Testing w the Indep-Measures t expensive Statistic • Split sample into many subsamples Hypothesis Testing w the Indep-Measures t Statistic • Only dif btwn t formula and z-score formula: ○ z-score uses actual pop variance σ or standard dev ○ t formula uses the corresponding sample variance s (or standard dev) when pop value is not known ○ To calculate indep samples t statistic: • For independent t, technically a test of dif btwn Determine dif btwn sample means & estimate sample mean difference & hypothesized pop mean error term (estimated SEM) dif  2 sources of error contributing to standard ○ Independent t: error in denominator ○ To estimate standard error term (s (M1-M2) need to know how to calculate pooled variance or (H0: 1 -2 = 0) pooled standard dev ○ Single sample:  Pooled Variance = s = Wpighted avrg of variance in each sample distr; works for both = & ≠ size samples • Same logic behind hypothesis testing applied as w ○ Assume samples come from pops w equal one sample variances ○ Error term typically defined as a s/√n or s /√n, • However, important difs w two grps: ○ Incorporate 2 pop means into hypotheses so easy to see how to use the pooled variance ○ Look at hypothetical distrs of sample mean or pooled standard deviation to estimate SEM 1. Make decision difs when testing hypotheses ○ So: Estimated standard error of mean (SEM) ○ If t statistic ratio indicates obtained dif btwn includes 2 measures of variability & 2 sample sample means (numerator) is substantially sizes greater than dif expected by chance (denominator) reject H , conclude there is a ○ Degrees of freedom include the n of each 0 group real mean dif btwn the two pops or treatments  Ex. df = (n 11) + (n 21) = df +1df = 2 1 ○ Often don’t use 1-tailed test since looks suspicious to increase chance of reject H 0 + n 2 2 st • To prepare data for analysis: 1 step = compute ○ Most 1-tailed tests would be equally valid as a sample mean & SS (or s or s ) for each of the two 2-tailed test samples  If don’t get what looking for w 1-tail, have • Can still use skewed data and make seem more to retest, can’t use same data w a 2-tail normal Hypothesis test follows same 4-step procedure in Ch. 8, 9 1. State hypotheses, select α level ○ For indep-measures test, H 0tates that there is no dif btwn the two pop means 2. Locate critical region. Critical values for t statistic are obtained using df (add together df of each Plotting Your Results sample) ○ Ex. N1=19, N 220  df = (19-1)+(20-1)=37 3. Compute the test statistic. The t statistic for the indep-measures design has same structure as single sample t in Ch. 9. ○ However, in indep-measures, all components of t formula are doubled: 2 sample means, 2 population means, & 2 sources of error contributing to standard error in denominator • Ex 1: N = 20 (observations per grp); N = 1 2 1048576 (observations per grp) – Ridiculously Large Sample (overpowered expt, P = 0) ○ d1= 0.027394031, d = 2.015805989 2 2 -5 ○ r1 = 0.000192419, r 2= 6.2473x10
More Less

Related notes for PSY201H1

Log In


OR

Join OneClass

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

Sign up

Join to view


OR

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.


Submit