Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
162 Cards in this Set
- Front
- Back
Name 3 requirments for parametric Analysis
|
1. subjects drawn from normally distrobuted population
2. groups demonstrate homogeniety of variance 3. interval or ratio data |
|
nonparametric test aka
nonparametric tests are based on... is are populations sometimes not normally distributed? |
distrobution free tests
ranking scores sample is too small |
|
3 criteria for choosing Non parametric analysis-
|
1. population is NOT normally distributed
2. No homogeniety of variance 3. nominal or ordinal data |
|
why dont we always use nonparametric tests?
|
1. power is lower!!
therefore we need more subjects to protect against a type two error. power efficacy = 65-95% 2. less of an ability to find a difference since you are comparing ranks |
|
What test do you use to check for two independant samples with a nonparametric test?
|
mann Whitney U-test
- VERY POWERFUL |
|
Relate Whitneys U to the t
How do you know if theres a difference when you run a U test? |
t- test = parametric
same as u-test = non parametric compare the R's The farther apart they are, no more different they are |
|
If the groups are of unequal size, group 1 sho9uld be the smaller group. The value for U is the ________ of U1 and U2
why? |
smaller
Always look at the lowest number because the closet U is to zero, the more DIFFERENT the groups are!! |
|
For t-test if t ___ Tcrit you reject the null.
For U- test if U_______ Ucritical then reject the null |
t > T reject
u < U reject |
|
when the sample size gets to ____ we decide sample is large enough to do parametric stats
|
25
|
|
Sign test and Wilcox signed Ranks are for .....
|
correlated (paried) samples
|
|
Sign test is good for what type of data
what is the sign based on? |
bimonial (yes/no)
ordinal (like MMT testing) direction of the difference x= # of fewer signs n= # of differences, ties dont count |
|
What does Wilcoxen Signed-Ranks test take into account
|
The direction and the magnitude of the differences. If the data is ordinal this test has more POWERRRRR
|
|
How do you know if the data is ordinal (if you are gonna do a wilcoxen sign test?
T = |
differences of two levels are greater than differences of one level
sum of ranks with least freq sign (total score) |
|
List how to do a Wilcoxen Signed Rank Test-
|
1. calcuate the differences between each subkect row
2. rank them 3. Give each rank the appropriate sign 4. determine which directino of change is the ;east frequent and sum the ranks for that direction. 5. T ? T critical if T is less then Tcrit then you reject the NULLy |
|
What do you do if you are using nonparametric data and you either have more than two groups or more than two levels of independent variables?
|
use a nonparametric ANOVA
called 1. Kruskal wallis One way analysis of variance of ranks 2. Friedman two-way analysis of variance by ranks |
|
What is the post hoc most often used for Kruskal Wallis one way analysis of variance by ranks?
|
Mann-Whitnet U-test with Bonferroni correction.
|
|
When would you use Freidmans Two way ranks anova?
|
for correlated measures designs (paired data- within subjects)
|
|
If you had 2 levels of one independant variable and independant groups what do you do if....
1 = parametric data 2 = non parametric |
1. do unpaired t test
2. U test |
|
If you had 2 levels of 1 independant variable and paired data....
1. and you had parametric data 2. And you had non parametric data (nominal ) 3. and you had nonparametric data (ordinal) |
1. paired t test
2. sign test 3. U test |
|
What is Students T test used for?
|
to compare two independant groups of subjects and assumes that all individual differences are distributed evenly between the two groups. (meaning, groups are evuivalent before the experiment)
|
|
After the treatment for the t-test, we test to see if....
what data can you use for the t-test (out of NOIR) |
a) if the two samples are still from the same population
b) or if the samples now represent different populations --> the treatment worked! meaning any difference is attributed to the treatment. Interval or ratio |
|
How does variablity and the means relate when doing a t -test?
|
no Variability and means are different = we are SURE the two groups are different
Variablity and means are different = we dont know WHY the means are different |
|
Directional or nondirectional
1. is a better than b? 2. does this treatment increase function? 3. Which is better, a or b? |
1. directional
2. directional 3. non directional |
|
When you plot a t-test on a graph, what can you tell by the overlap?
what if the overlap in the distrobution is greater to the left than the distrobution to the right? How do we know how big the difference has to be before we can be sure that its due to treatment? |
depending on the amount of overlap in the twl distrobutions, there is a greater or lesser likelihood that the treatment is causing the difference between the two groups.
( so you can eyeball these stats) the difference in the distrobutions to the left is more likly to be due to variablity in the overal populations and no to treatment effect than is the difference in the distributions to the right. statistical ratio |
|
variability within groups means-
|
estimation of sampling error
|
|
WRT a t-test, if the difference between groups is due to sampling error (meaning no treatment effect) then...
|
the statistical ratio witll be 1
if there is a treatment effect the statistical ratio will be greater than 1 statistical ration = SR |
|
independant samples means
|
Unpaired
|
|
What assumptions do you have for Independant sample t test?
|
1. independant groups
2. RA 3. normal distrobution 4. variance of the two groups is relatively equal |
|
what 3 things do we need to know for appendix A.2 (to find critical value of t)
|
1. power
2. one or 2 tailed test 3. DOF 2 tail a No=b =nondirectional 1 tail a >b = i direction |
|
Degrees of Freedom means?
|
statiscial concept indicating the number of values within a distrobution that are free to vary, given restrictions on the dataset; usually N-1
|
|
Problems with t-test when unequal variance...
|
1. if you have inequal sample sizes and the larger sample has the larger variance, t test is less powerful
2. if you have unequal sample sizes and the larger sample has the smaller variance then the priobability of a type 1 error is increased *** especially if the variance of the one group is more than twice the variance of the other group. |
|
So running a two tailed test would make it a little easier to reject the null hypothesis since power is 2.5% insterad of 5%
|
get it got it great!
|
|
How can you do a paired t test but not necessarily have to do repeated measures?
|
you have have matched pairs of subjects that are pout into two groups, but are compared to each other one-on-one
|
|
so what do you care about if you do a paired t-test?
|
Only how each of the individuals changed with treatment. We no longer care about the absolute score.
|
|
df for a paired test =
|
number of pairs- 1
|
|
The more comparisons you make the greater probabiltiiy of a type 1 error so.....
|
dont do multiple t tests you crazy!
|
|
what would you do instead of multiple t-tests? (meaNING IF YOU HAVE more then 2 groups)
|
run an anova
also FYI- a two way anova is a factorial design! cooolio |
|
ANOVA=
simplest anova called--- when would you use it? what would you say if you rejected the null hpyothesis with this test? so then what? |
Analysis of variance
anova for independant samples: one way classification (one way anova) for an experiemtn with one single factor (independant variable) with three or more levels. So we are comparing 3 group means. There is a difference between a, b, and c. But we dont know WHERE the difference is then calculate the statistical ratio. then run post hoc |
|
for more than two groups we use __________ to represent the variablity
|
sum of squares
|
|
between group sum of sqaures =
|
how far is the group mean from the grand mean (meaures of variability between groups.
|
|
post hoc aka-
|
multiple comparison testt
|
|
factorial design =
what do they allow you to do? |
(2 way anova)
two independant groups two independant variables with levels of each find an interaction between things. |
|
what would you run with: one subject design, one group tested under different conditions-
|
repeated measures anova
|
|
If your groups have unequal variance what should you do?
|
correct for this using either greenhouse-geisser epsilon or huynh feldt epislon yahhhhhh
|
|
Probability=
probability predicts... |
the liklihood of any one event occuring, given all possible outcomes.
what should happen over the long run NOT for any given trial or event !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! |
|
why do we care about probability?
|
inferencial stats are based on describing the probability that the results obtained are due to chance.
|
|
Sampling Error=
|
estimation of population parameters from sample statistics in based on the assumption that the sample is randomly drawm from the population. and validly represents the population.
|
|
can we know the population parameter in most cases?
|
NOOO so we need a better way of estimating this... like Standard error.. oooooo
|
|
Standard Error=
calculated from.. used to claculate the... |
the Standard error of the mean is an estimate of the population SD.
theoretical examples. confidence interval |
|
Standard error of the mean =
so what percent of the means falls within +/- 1 SD? |
is the SD of the distrobution of means of randomly drawn samples from a population.
68% |
|
Instead of a point estimation of the population mean we can use an (n)_____ estimate
|
interval
|
|
Confidence Interval =
the size of the confidence interval is based on.... as sample size increases, the interval.... |
is a range of scores with speciic boundaries that probably contains the population mean
the sample mean and the standard error. The more confident we wish to be that the interval contains the mean, the larger the interval will be. decreases (gets smaller) |
|
discribe confidene interval for normal distrobution
|
with in +/- 1.96 of the SD
95% |
|
W/r/t confidence intervals, when you have a large sample size use ____ instead of ____
|
t-distrobution instead of normal distrobution. You end up using a bigger number than 1.96, makingyour CI's bigger
|
|
Whats better to express than the SD?
|
CI
|
|
Type I error vs Type II
|
comission vs omission
(wrong choice) vs (should've) alpha vs beta |
|
List ways to Increase Power!
|
1. inc sample size
2. increase effect size 3. dec variance |
|
Why should you dec variance to inc power?
|
TO do it you use a sample thats less different.
and decrease it because it the less different it is, the less you can generalize about it |
|
1/two tailed vs directions
|
1 tail = directional
2 tail = non directional (more conservative) |
|
level of significane =
the bigger the alpha the ______ chance that youre wrong |
predetermined probability of making a type I error (alpha is chosen a priori)
bigger the chance that youre wrong |
|
If you have a reason to not want to make a type I error, then make alpha.....
|
even smaller
|
|
Statistical Power =
|
(1-B) is the probabilityu of failing to reject a false null hypothesis.
The complement of this is power. (B-1) BTW- if you inc alpha and dec beta you INC POWER |
|
ppppPOWER =
usual values of power = |
The probability that a test will lead to rejection of the null hyp[othesis if there is a real difference
80-90% |
|
Power depends on 4 things
how does it help you reject the null? |
1. significant criterion
2. variance in the data 3. sample size 4. effect size p<alpha reject null! |
|
How can you increase the effect size
|
-treat longer period of time
-choose population who's before scores are worse |
|
THESE THINGS YOU SHOULD NOOOOOOOT SAY
|
"EXTREMELY SIGNIFACANT"
or change your alpha so the study looks more conclusive |
|
if you flip a coin 10 times and get heads, whats the probabilyt of getting heads on the 11th?
|
50%
|
|
descriptive statistics =
Inferenctial statistics = |
describe the data literally
infer somthing from data to a larger group (helps us generalize) |
|
Whats first thing you do with descriptive data?
then |
put it in order. ( like put everyones midterm grades in order)
then create a frequency histogram, and read the shape |
|
Sometimes its easier to look at the overal shape of a distribution by using a _________
|
larger BIN
|
|
a distribution is skewed in the direction of ....
|
the tail!
|
|
with descriptive stats
1. Mean > median 2. Mean < median 3. mean=median=mode |
1. skewed RIGHT
2. skewed LEFT 3. normal distribution |
|
descriptive test fails when....
|
either it comes out bimodal or sample size was too small to get accurate distribution of whatever....
|
|
Variability =
measures of variability |
a description of how the scores are distributed about the mean
RANGE: the difference between the highest and lowest score. |
|
The range (measure of variance) is very sensitive to...
and that strongly effects... |
outliers
and outliers strongly effect the mean |
|
To avoid outliers (by using the mean) use...
|
percentiles
|
|
Percentiles =
|
describe relative position of data in a distribution. turns ratio data into ordinal data!
(good for sat scores but not much else) |
|
Variance=
since variance is a pain in the ass to think about.... |
describes how far each score is from the mean score. Variance describes the spread of data.
we use SD |
|
When do you use the following...
1. range = 2. variance = |
1. if no outliers (general discription)
2. statistician, interval data |
|
SD =
how do you know what units of measurement to use? |
the square root of the variance.
same units of measre as does the parameter being measured. |
|
Coefficient of Variation =
name 2 advantages- |
standard deviation expressed as a percentage of the mean
1. uniltess 2. describes variability as a proportion of the mean- relative variance. |
|
usually to describe data, you use the range and ____
what should you use if you want to express variation as a percentage? |
either variance, SD , or coeff of variance.
use coeffcient of variance |
|
Naturalistic research design =
|
qualitative research
|
|
ontology =
|
branch of metaphysics dealing weith the natire of being, reality or ultimate substance. Ontology asks- what is reality-
|
|
Epistemology=
|
the study or theory of the origin, nature, methods and limits of knowledge. "how do we come to know reality"
|
|
Logical positivism (d. Hume) =
|
the world is objectivley knowable and can be discrovered through observation and measurement that is considered unbaise
|
|
what is the basis of quantitative experimental research?
quantitative experimentalism aka - |
Logical positivism- encompasses the belief that it is possible to know and understand phenomena that reside outside of us. Only through observation and sense data can we come to know truth and reality
empiricism, rationalistic or posivist. |
|
bottom line difference between logical positivism and hollisit cperspective=
|
there is one objective truth, VS truth is different, depending on the person
|
|
Qualitative paradigm states:
|
individuas create their own subjective realities and thus the knower and the lowledge are interrelated and interdependant
|
|
Assumptions of Qualitative/ naturalistic Paradigm
|
1. multiple constructed realities
2. interdependace of researcher and subject 3. knowledge is time and context dependant 4. cant distinguish casue from effect 5. resaerch is value bound |
|
what does multiple constructed realities mean?
|
- no single truth
- each person has own reality and attach meaning to the events AND actually creates their own reality |
|
what does interdependance of researcher and subject mean?
|
- because you are studying something you can change that which you are studying
- hawthorne effect - cant study something without changing it |
|
What does it mean for researches to believe that you cant distinguish cause from effect
|
you can only describe relationship between events and behaviors !!!!!!!!!!!!!!!!!!!!!!!! basic to quantitative reserach-
|
|
value-bound research?
|
- there is no such thing as objectivity
- all researchers are biased - no researcher is disspassionate - the values of the researchers efgfect the outcome of the research |
|
Name 4 naturalistic inquirys essential characteristics
|
1. human experience cannot be understood by reductionism
2. meaning in human experience is derives from an understanding of individuals in their social envirenments 3. multiple reailites exsits 4. those who have the experiences are the most knowleaged about them |
|
Quant research call the people in their study....
Quals call them... |
subjects
participants |
|
qualitative approach =
|
phenomonology
|
|
Ethnography=
|
the study of the social milieu of a specific cultural group of people. Examines the attitudes, belifs, and behaviors of that culture.
|
|
Grounded Theory
|
systemic discovery of theory from the data of social research. a method used with this stratedgy is the conactnt comparison method
|
|
Endogenous Research-
|
research that is conceptualized, designed, and conducted by researchers who are insiders of a culture, using thei won epistemology and their won structure of relevence.
|
|
Participation Action research=
|
similair to endogenous, its conceptualized, designed and conducted by researchers who are insiders of a culture, however the purpose is to generate knowledge to inform action.
|
|
In health care the most common mthodologies are...
|
ethnograophy, phenomenology and grounded theory.
|
|
what if we want to conpare interval or ratio data to so if they are different?
what if the data is nominal or ordinal? |
run a t test
analysis of frequencies |
|
How would a quantitative person and and qualitative person do the following study:
"How does a 10 week program affect balance in individuals with chronic strokes" |
quant: 2 group control randomized design. assumes one reality
QUAL: (phenomenology) recruit people with chronic strokes and listen to them about what balance means to the,. Assign program for their needs. |
|
List 5 ways (perspectivces) to approach a a question in Qualitative Research
|
1. phenomenlogy (human meaning can only be understood through experience)
2. Ethnography (examines beliefs, behavoirs, of a specific culture) And a culture could be like people with wheelchairs. 3. grounded theory (constant comparison method) 4. Endogenous Research (insiders of a culture can do the experiment) 5. Participatory action research - research done be insiders to decide what to do and why. |
|
Qualitative researchers have two methods to collect samples-
|
1. non-probabilistic(non random)
- usually conveient - purposfuol or snowball (friends....) 2. Informants are sometimes recruited as you go along. |
|
The two most common methods of data collection (I think for qual is..)
|
1. interviews
2. observation interviews are direct contact between researcher and participant in the participants natural envirnment. quant's collect numbers while qual's coolect words |
|
structured =
semistructured = Unstructred = |
administration of questionaire
some initial questions but with some flexibility to follow where the partcipant leads general topic but let the interview flow. |
|
qualitative researchers call their data-
|
RICH - wow there is so much fucking information here holy shit!
|
|
what types of obsrevation does the QAUL researcher have to pic froM?
|
direct observation = allows the researcher to understand the context of the incident being discribed.
complete participatation participant as observer observer as participant complete observer |
|
participant as observer =
|
the researcher assumes limited membership in the group (this may or may not be disclosed to the participants)
|
|
observer as participant=
|
there is no membership in the group, only brief contact with group and the researcher is still considered to be effecting the behavior of the group since the group knows they are being watched
|
|
complete observer=
|
objective observer- the participants may not even be aware of the fact that they may be affecting the behavior of the participants.
|
|
Artifacts=
|
material traces - physical evidence
|
|
Artifacts=
|
material traces - physical evidence
|
|
examples of artifacts
|
written documents
records - office writings objects - anything!! |
|
Discribe quaLitative data analysis-
|
tons of it!
content analysis is where they search for common themes that repeat in the discussion these themes are coded and then identified throughout the data. And maybe will create models to discribe what they found |
|
If the reliability of the QUAL is all subjective, how can they determine reliability and validity?
|
1. triangulation
2. re-coding 3. member checking 4. audit trail |
|
triangulation =
re-coding = member checking= Audit trail= |
using multiple scores of data to confirm concepts
having a second person code the data- did they recognize the same things? after a researcher has drawn some conclusions about what they have observed, check with the membrs of group and see if they agree a diary of everything you did. |
|
qualitative research is _____. This process isnt linear
|
Irative (read over and over and over again.)
1. collect data 2. analyze data 3. report results usualy... 1. collect data 2. begin analyzis 3. recognize themes 4. repeat steps |
|
How does a qualitative researcher know that they are done?
_______ research is inductive. Its hypothesis generating... |
SATURATION - you are no longer discovering anything new.
quaLitative research - and people use it when there isnt enough info to design a real study, lol. |
|
All stats so far have looked up to compare one independant vaiable across 2 groups or two measurements of the same groups.
What if we measure 2 or more variables? and wwant to see how these parameters change with respect to each other? |
do a correlation
|
|
how do you know if your correlation is strong?
|
calculate a correlation coeficient - this discribes the magnitude and direction of a correlation.
|
|
Whats the first thing you need to do with correlation data?
|
PLOt the data- spo you can pick out the pattern an dknow which type of correlation to run
|
|
correlation magnitude =
|
1.00 is perfect
0.00 is NO correlation 1.00- looks like all dots on the same line |
|
Discribe "direction" w/r/t a correlation
R = R2 = |
+ as X inc so does Y
- as X inc, Y dec R = correlation coeff. helps you predict things by giving strength and direction determinations? |
|
Correlations
.25 = .50 = .75 = +.75 = |
little/non
fair mod great |
|
Intercorrelation of more than two different variables can be studied at the same tine with a
|
correlation mix
(this matrix chart discribes the relationships at once) |
|
A correlation can be considered significant if p __ p crit
Correlations are very sensitive too... |
p < pcrit
sample size (so small samples sizes can lead to significant correlations) |
|
Which correlation to run..
1. ratio data 2. Ordinal data 3. Dichotomous Variable 4. 1 dichotomous, 2 continues |
1. (r) pearson product moment correlation coeff (measures how far each one is from a point
2. Spearman Rank correlation Coeff 3. PHI coeff 4. point diserial or rank biserial |
|
What does Correlation NOT mean?
|
agreement
causation |
|
When you run a linear correlation, you fit a line into the data. The corrleation coeff tells you how well the data fits the line.
The equasion of this line is called- |
regression equasion
and can be used to predict values of Y |
|
Coefficient of determination =
|
R2
indicates the % of the total variance in the y scores that is accounted for by the X scores. The fraction of the information you would need to accuratle predit the Y score given the X score |
|
With a correlation of r = .7 you still only have r2 = .49 or ...
|
less then half of the variablity in Y that can be explained by X
|
|
whats the outcome of logistic progression?
|
y/n variable
|
|
Outliers
|
- rules of thumb > 3SD from mean
- case by case basis - watch for discarded data in research reports |
|
what if we want to compare 2 exams o see if they are different?
|
if they are interval or ratio data we run a T test.
if they are ordinal or nominal data we do analysis of variance |
|
What is evaluted with X2?
|
comparison of porportion of answers observed within a distrobution and the proportion expected by chance
|
|
2 Chi Square Assumptions-
If the observed minus the exected difference inc, the probability that... |
1. freq reprsent individual accounts, not %'s
2. catagories are exhaustive and mutually exclusive there is a difference between their results gets bigger (um duh?) |
|
If X2 < X2 critical....
|
we FAIL TO REJECT THE NULL
|
|
If we dont know the "expected" probability of a question we..
|
assume its uniform so we assume the answer is yes (that you calculated from your data score)
This means assume a uniform distrobutiion! |
|
So once you have your uniform distrobution you can compare to your distrobution?
|
arrange scores in == and non -overlappoing intervlas. set interval width = to the standard deviation of the scores.
at x and X+1 we expect to see 34.13% but whatever the hell you do in the end just compare X criticals to see if you should reject the null |
|
what if you had a circle all that apply - on your sheet, what test would you use for that correlation?
|
McNemar test (a variation of the chi squared)
|
|
Measurement =
|
true value + measurement Error
All measurements have some degree of error in them. Nothing can be measured exactly! |
|
Two types of error =
which one effects reliability? |
systemic error and random error
RANDOM 0.0 = NO reliabitliy 1.0 = perfect reliability |
|
Effect of variance on reliability-
|
1. if group is too homo you will get a poor reliability coefficient even if the errors are small
2. unspoken truth = if a group of scores used to evaluate the reliability of a measure is too HETER, you will get a good reliability coefficient, even if the errors are large. |
|
Why is correlation not a good meaure of reliability?
|
even if things are highly correlated like hight and shoe size... your height does not equal your shoe size. pearsons correlation looks at covariance- the ranked order of mearuements in a dataset
ise intraclass correlation coeff (ICC) instead |
|
what is the ICC?
|
its a correlation coeiff claculated using variance estimates obtained from asn ANOVA - therefore it not only looks at correspondence but also at agreement amoung ratings.
|
|
Strengths of the ICC
|
1. can asses reliability amoung two or more raters
2. doesnt require same numbers of raters for each subject 3. designed for interval and ratio data 4. can be used with nominal data?? but if you have ordinal data use a KAppa instead |
|
models of ICC are distinguied by
|
how raters are chosen abd assigned to subjects. This also effects generalizability of results
|
|
discribe model 1 of ICC
|
each subject is assessed by a different set of k raters. These raters are selected randomly from a larger population.
(this is almost never used) |
|
Discribe model 2 of ICC
|
each subject is asses by each rater- these raters are selected randomly rfom a larger population
(good if you need to generalize) |
|
Discribe model 3 for ICC
which model is the most conservative method for reliabiltu (and rarely used in clinical reliability studies?) |
each subject is assess by each rater- but the raters represent the only repreasent the raters of interest. (like hired professional opinions)
used when you DONT need to generalize results. model 1 |
|
each of the three models of ICC have __ forms
what does it mean if you see ICC (3,5) ? |
2
one for individual ratings and one for mean ratings model three was used and the ratings are the average of 5 measurements. this isnt a standard notation so ton use it. use ICC(3, k) |
|
k =
n = how do you calculate ICC? |
number of raters
numnber of subjects tested run anove repeated measures (two factor without replication) |
|
How do you interpret the ICC?
|
there are no standards. Dr besser says... above.75 is good
|
|
What are 2 main reasons for low ICC's?
And what do you do? |
1. the raters dont agree
- so find out who was unreliable and fix it. - refine your methodology so everyone agrees on how to make the measurement. 2. the variablity amoung subjects was insufficient - test more subjects over a wider range of your dependant variable. but dont be biased. |
|
Name two other forms of agreenment (what ever that means w/r/t ICC)
|
1. % agreement - for nominal or ordinal data
PROB-> this doesnt take possibility of chance into account 2. Kappa Statistic- chance corrected measure of agreement for nominal or ordinal data - a better representation of reliability |
|
On a Kappa cell, which is the best ones to be in?
which are the worst ones to be in? |
the ones with the same name like inde and inde, assiss and assis and depen and depen
the inde and dep (both of them) |
|
Whats a good k ?
|
greater or equal to 80% is excellent. then to 60% is substantial. Then to 40% is moderate and below that is poor
*highly dependant on what you are going to do with the data* |
|
Whats a weighted k?
|
you can penalize raters more for being farther apart.
|
|
how do you evaluate internal consistency for ICC and Kappa?
|
chronbachs alpha
- to see if items were messaured by the same construct. |