evaluating psych
Terms
undefined, object
copy deck
- type II error
-
may fail to reject Ho when it is false
-BETA LEVEL:prob type II error (.20 or 20%)more willing to accept - power
-
degree to which research design sensitive to effects of IV-detect effects IV more easily than less powerful designs
- #participants in study
-< # < detect effects IV on DV(magnitude) - power analysis
-
determine # participants needed order detect effect IV
-aim exceed .80 - effect size
-
proportion variability in DV that is due to IV
-.00 to 1.00 - T-test
- analyze data from 2 group randomized exper./its error/variance we see random?
- 1. directional hypoth and 2. non directional hypoth
-
1.states which of 2 conditions means is expected to be larger
2.states 2 means are expected to differ, no prediction which is larger - 1.one tailed test vs. 2.two tailed test
-
1.used to test directional hypoth.
2.hypoth is non-directional - paired t-test
-
used when exper. involves matched subj. or w/ in subj. design
-participants 2 conditions similar if not identical related to DV -
paired t-test
1.matched subj design
2.w/ in subj design -
1.randomly assign matched pairs to two conditions
-score up DV in one then score up DV other situation
2.same participants serve in both conditions
- (+) corr. obtained btw. matched scores - stat. significance affected by 3 factors
-
1.SAMPLE SIZE:larger size more stat sig.
2.MAGNITUDE CORR:larger value r less likely be .00
3.NOT DRAW IN CORRECT CONCLUSIONS:sig. diff from 0 if there is >5% chance corr. as large as one obtained could have come from pop. w/ true corr. zero
->.16 corr. could be 0
-.10 weak magnitude/.30 moderate/.50 strong - deonotology
- ethical approach maintaining right and wrong be judged according to universal moral code
- ethical skepticism
-
denies exist concrete and inviolate moral codes
-relative culture and time - utilitarian
-
judgement action depend on consequences of that action
-APA uses this - cost-benefit analysis
- weigh pros and cons fo study
- benefits behavioral research
- when cost benefit analysis is conducted
- basic knowledge
- up risk and cost ok when research up
- improve research/assessment techniques
- moe research like this- improve research itself/ more reliable, valid,useful/efficient methods
- practical outcomes
-
up welfare subjects
-better drugs/treatments - benefits researchers
- education function- first hand research
- cost
-
-balance risk and cost
-$ involved conduct actual research
-deception=distrust toward research - balance benefits and cost
-
down benefit down cost and vice versa
-research approved by IRB - informed concent
-
-inform nature study and agree to participate
-inform about features research influencce willingness to participate
-pain discomfort
-IC form usually used
-IC given orally some cases only if witness present - ?IC
-
-COMPROMISING VALIDITY STUDY-act diff if watched
-UNABLE GIVE IC-child, mental handicapped/ parent/legal guradian give IC
-GREATER BURDEN OBTAIN IT THAN NOT-
-intersection street!
-min. risk, not affect rights &welfare, not carried out if required - invasion privacy
-
-no strict guidelines
-study w/ out person knowledge judgements left to investigator - coercion participate
-
prof. require students participate research studies
-paper option=research time lab - physical and mental stress
-
-difficult study when expose small amounts phy. mental stress
-lead beleive dying=NO!!
-decisions made cost-benefit analysis research - deception research
-
-use confederate
-false feedback
-related studies as unrelated
-incorrect info reguarding stimuli materials - objections deception
-
-lying and deciet immoral?
-ends justify means
-violation moral rules
-widely used enter knowing this suspicious researcher tell them right away
-distrust behavioral scientist & research in gen.
-participants do not view decep. as same as lying! informed after ok!! - debriefing
-
-when deception used must use this!
1. clairfy study- explain reasons deception occured
2. remove excess stress or (-)consequences
3. obtain participants reactions/feedback on study
4. leave feeling good about study - confidentiality
-
-only purpose for research
-ensure responses anonymous
-give codes label data
ex:milgram exper= obedience - reserach with animals
- -same as humans only different guidelines to follow
- scientific misconduct
-
-ethical standards memb. guard against this
1. fabrication, falisfication, and plagiarism
2. fail report data not with their own views or making data availiable who wish to varify their data
3.research participants that are problematic (what order authors go)
4. sexual harrasment/abuse power, discriminate, or failure follow gov. regulations
5. conduct poorly designed research - factors distort corr. coefficients
-
1.RESTRICTION RANGE:narrow rep. data
-misleading or weak corr?
2.OUTLIERS:deviant from remainder of data
-further 3 SD away from mean
3.RELIABILITY MEASURES:>reliable measure, down corr. coeff. will obtain - corr. and causality
-
-CORR. DOES NOT IMPLY CAUSALITY!!
-even if perfectly corr.
-basis tobacco company claim against cigs and lung cancer - partial corr.
-
-can conclude particular causal explaination of relatiohship btwn variables more likely be correct than are other causal explainations
-x and y corr
=x may cause y/vice versa/some ther var. cause x and y
-corr. btwn variables stat. removed
-eliminate 3rd variable possible influnece
-see if x and y stil corr. - analyzing experimental data
-
-diff. group ave. suggest IV had an effect
-can not draw conclusions about effects IV simply by looking only at means of experimental conditions - cross-sectional survey
-
single group
-char. that group/various groups differ - sucessive indep. samples survey
-
2 or more samples of respondents answer same ? at different pts. in time
-validity on if each selected same way - longitudinal
-
single group/ long time
-changes in behavior studied - demographics
-
-describe patterns basic life events and experiences
-birth, marriage, employ, migration, death - epidemiological research
-
-study occurances disease diff. groups people
-info groups at risk illness/injury - surveys
- all pic. how large groups tend think, feel and behave
- sampling
- -way select people for study from population of interest
- representative sample
- one which can draq accurate, unbiased est. char. larger pop.
- sampling error
- cause reseults obtained sample diff. entire population
- error estm.
-
(margin error) degree data obtained sample expected deviate from pop. as whole
-condfidence expressed in term error estm.
-down error estm. more close resuls from sample estm behavior large pop.
-function sample size, pop size, variance data - economic sample
-
reasonably accurate estm. of pop. reasonable ost and effort
-error estm down when pop. up
-larger variance up sample must be - probability sample
- prop and indiv in pop. will be included
- epsem design
-
specifies all cases in pop have= prob. being chosen for sample
-"equal prob. selection method" - simple random sampling
- every possible sample desired size same chance being selected from pop.
- sampling frame
- list pop. sample comes from
- table random #'s
- use with large pop
- stratified random sampling
- variation SRS divide pop. into 2 or more strata
- stratum
-
-subset pop. shares particular char.
-compare 2 group with in this - cluster samp
- sample group participant based naturally occuring groups/get list/randomly select list from that group
- multi stage process
-
-no sample frame needed
-less time/ effort needed contact participants - error variance cause mean diff
-
means may differ even if IV does not have an effect
-CONFOUNDING VAR:portion total variance in set scores due to extraneous variable that differ systematically between experimental groups (2ndary variance)
-error variance in data ave recall 2 groups of participants likely differ slightly even if treated the same
-indiv. differences - nonresponse problem
-
-failure obtain responses indiv researches select from sample
-data may be bias was impossible determine
-representativeness of sample compromised
-same with tele surveys
1.take steps to increase response rate
2.if respondents and nonrespondents differ in any systematic ways - non prob samples
-
-no way knowing prob. a particular case will be choosen for sample
-not calculate error estm.
-limit ability generalize
-test how variable relate to research test hypoth.
-external validity assessed through replication
-sim. findings validity up - convenience sampling
-
readily avaliable used
-prob. using college students
-more intell. than gen. pop. - quota sampling
- certian kind participants obtained particular proportions
- purpose sampling
-
use judgements decide which participants include in sample
-try get respondents typical of pop.
-elections use this -
describing and presenting data
critera good one -
-accuracy, conciseness, and understandability
-RAW DATA-all participants scores on all measures
-selecitve data present=clearly describe results
-description understood - numerical methods
- sum data from # as % or means
- graphical methods
- present data graphically or pictorial form (graphs)
- frequency distributions
- table sum raw data # scores fall with in each several categories
- simple frequency distrib.
-
# participants who obtained each score
-up or down
-2nd column frequency -
inferential stats
(solution to error variance cause mean diff) -
-conclude independ. var. has effect when diff. btw means experimental conditions larger we expect to be when diff due solely to effects of error variance
-compare diff btwn means of exper. condition of diff. expect to obtain based on error var.
-never know for sure
-can specify probability - group frequency distrib.
-
-frequency subset of scores
-break range scores several subsets(class interval)=size - hypothesis testing
-
Ho:IV no effect on DV
H1:IV did have effect of DV
REJECT Ho:IV did have effect on DV
FAIL TO REJECT:IV did not have effect of DV - relative freq.
-
proportion total # scores tells each class interval freq. class interval % total # scores
-class interval are ME
-capture all possible reponses
-all class interval same size
-freq. info=histogram/bargraphs/ freq. polygon - type I error
-
researcher concludes IV had effec on DV when the observed diff. btwn means of experimental conditions actually due to error variance
-ALPHA LEVEL:prob. making type 1 error
-.05 or 5% (>.05 reject)
-STAT SIGNIF:when reject Ho with low prob. making type 1 error refer diff. btw means - measures central tendency
-
convey info. about distrib. providing info about ave/ most typical
MEAN:ave (can be misleading)
MEDIAN:middle
MODE:most often - measure variablility
-
-know how much scores distrib. vary
-descriptive stat convey info. spread/ var. set data
-RANGE:least useful set on two extremes
-VARIANCE:takes into account all of scores when calc. variability of set of data
-SD= it's square root
-index ave. amount variability in set data
-purpose stat. analysis - z-score
-
particular participants score relative to rest of data (standard score)
z=(y1-y)/s -
SD and normal curve
normal distribution -
-having mean and SD estm. percent participants who obtained certian scores
-tell distrib. across various ranges of scores
-useful identify extremes/outliers
-very high or very low z-score
-(-3.00 or +3.00) - sucessive independent
-
cross-section with time component
different people/repeatedly - SD
-
I 34.13%-68%
II 13.6%-94%
III 2.14%-98%
IV .13%-99% - corr. research
-
whether var. are related to one another
-relatiohship btw. 2 or more naturally occuring variables - Pearson CC
-
(r)
-interp. consider 1.sign and 2.magnitude
1.direction (+) positive relationship(up and up)
-optimism and health corr.
-(-)=inverse or negative relationship btw 2 var (up and down)
2.numerical value ignore sign
-strength relation between variables
-.00 not linerly related - corr. coeficients
-
-1 to +1
-strength and direction relationship
-only requirement obtain scores on 2 var. for each participant in our sample - 1.scatter plot, 2.perfect corr, 3.zero corr.
-
1.graph representation participants scores 2 variables
2.(-1.00 to +1.00)
3.no pattern - coefficient of determination
- x2 of corr. coefficient; indicated proportion of variance 1 var. can be accounted for by other var.
-
1.variance
2.systematic variance
3.error variance -
1.amount var. in set data
2.part total var. in participants responses related var. researcher is investigating
3.portion total variance unrelated to variables under investigation - stat. significance of r
-
-corr. coeff. calculated on a sample has very low prob. being zero in pop.
-b/c sampling error never get r=.00 even though true corr. pop is zero
->.05 stat signif. -
Joe Brady:
executive monkeys experiment -
Type II error
monkeys first group moved on to the second group only if did good in phaseI
-actually looked at two diff populations of monkeyes
-confounding variables
-found monkey A (the boss) had more ulcers - Weis
-
same study as brady but found opposite
-monkey b more than monkey A got ulcers -
Jarrard:
LSD experiment -
rats press for food (motivation)
problem results:
could be a build up of LSD in system cause drastic decline btwn .2 and .4
-better:take 6 diff rats and give them each dose (use latin square) - counter balancing
-
procedure used within subj. designs in which diff participants recieve the levels of IV in different orders
-used to avoid systematic order effects -
btw group:
1.pros and 2.cons -
1. faster, someone drops out find another just like them, everyone diff indiv impact
2. more $, less power, very difficult match subjects -
w/ in groups:
pros and cons -
1. more power, every person own control group, removes error from indiv variability
2. longer, slower, when someone drops out problematic - 1.carryover effect and 2.order effect
-
1.a situation in w/in subj. designs in which the effects of one level of the IV are still present when another level of IV is introduced
2.an effect on behavior produced by specific order in which levels of the IV are administered in a w/in subj design -
Greene 1996 Memory Test
study list->delay (5sec and 20sec)
*do both with in and between subj. designs* -
results: with in better becasue found diff where as with between found no difference
WITH IN BETTER!!!!! -
counterbalancing:
complete and incomplete -
complete= what lit for levels?
3 conditions
incomplete= latin squares
doesnt count for all possibilities