MCRS LECTURES 5-10

units of analysis/cases subjects or objects that are mentioned (respondents, newspapers etc)

variables characteristics of units of analysis

values possible categories per characteristics (gender, age, education level etc)

measurable definition of a concept description of the meaning of a concept in research that is as accurate as possible

concept construct or variable

manifest (measured variable) can be directly measured or observed

latent variable variable that is not directly observed (requires multiple definitions to be measured)

reliability consistency (is it replicable)

validity is it right?

range maximum value minus minimum value

interquartile range "middle ground of dispersion" ignoring top and bottom 25%

Standard Deviation (SD) square root of variable

why do we use SD to make generalizations about a wider population from which we've drawn our sample and to calculate probabilty of any particular result occuring

Variance step 1 subtract mean value for the group from each individual value

variance step 2 square each result (to eliminate problem of dealing with negative number)

variance step 3 & 4 add the results , divide he sum of squares by number of values minus 1 to get an average of the squared variations from the mean

z-score number of units of SD any one value is above or below the mean

why do we use z-scores it allows us to compare numbers from different measuring systems

z-score interpretation the larger the z-score, the further its value from the groups mean and vice versa

nominal measurement and methods Mode (pie chart, bar graph, frequency table)

ordinal measurement and methods mode, median (bar graph, frequency table)

interval measurement and method mode, median, MEAN (histogram)

ratio measurement and method mode, median, MEAN (histogram)

central tendency mode mean median

normal distribution condition 1 always bell-shaped

normal distribution condition 2 sample size should be at least 100

example of normal distribution exam grades, age, income

skewness left or right skew. adds extremescores or outliers to the sample. makes mean and median different from a normal distribution

what measure of central tendency if distribution is skewed? median

negative value on skewness left-skewed

positive value on skewness right skew

nominal variable dispersion no dispersion

at least ordinal variable dispersion range

at least interval variable dispersion standard deviation, variance, deviance (error)

dispersion how different is each score from the center of a distribution (arithmetic mean)

sample group where you get your data from

deviance (error) negative value below the mean minus mean score (x^i - xBar)

sum of squared errors (ss) indicates total dispersion or total deviance of scores from the mean

average dispersion variance

normal distribution percentage 2,5% 13,5% 34% 34% 13,5% 2,5%

why do we use standard normal distribution to visualise data in same type of distribution (through z-scores)

correlations are (1) a way of measuring the extent to which 2 variables are related, a measure of the degree of association (symmetry) among variables

correlations are (2) indicates if variable changes in predictable manner as another variable changes, examines if one increases, the other also increases/decreases/stays the same

pearson product moment correlation (Pearsons R) degree of association between 2 scale (interval, ratio) variables

covariance when 2 variables covary. knowing how one variable changes helps in predicting how another variable changes

covariance interpretation when variables covary, they are related to some extent (correlation among variables)

-1 or +1 correlation perfect correlation

<.3 correlation weak

.3-.5 correlation moderate

>.5 correlation strong

coefficient of determination squaring one value of r you get the proportion of variance in one variable shared by the other

r2 coefficient of determination

sign of correlation (+ or -) indicates direction of relationship

exploratory factor analysis (data reduction technique) measure latent constructs, search for patterns (dimensions/factors), establish one or more factors/dimensions

what kind of data is correlation valid for numerical data

1 item measured on a 5-point scale use ordinal categorization

1 item on a 7-point scale use interval categorization

2 or more items measured on 5/7 point scale use interval categorization (if true zero)

validity definition accuracy of the measure

reliability definition consistency of the measure

coefficient is the same as? communalities

common variance variance that a variable shares with other variables

unique variance variance that is unique to the particular variable

common cause among variables indication for latent factor

uninteresting factor loadings >.3 eigen value

multiple regression 2 independent variables and 2 dependent variables measured on interval level

simple regression 1 scale is used

what if there are 2 dependent variables 2 simple regression analyses

what rotation method do we use? oblique (direct oblimin)

when do we use direct oblimin to rotate factors? when we expect a relationship between factors

steps for factor analysis 1. EFA 2. Crohnbach's alpha, check if recoding negatives is necessary 3. Create variable/scale/composite mean (compute variable) 4. Descriptives (mean and sd)

Correlation is not causality

this variable explains the relationship between two variables mediator variable

non-spuriousness nothing except X must be influencing Y

goal of measurement aim for highest level of measurement

categorical measurement nominal/ordinal (cannot calculate categories or ranking)

numerical (continuous) ratio/interval

crohnbachs alpha reliability analysis, (important to reverse code negative variables)

why do we use factor rotation because we always expect underlying dimensions of a latent construct to correlate, which is why we use oblique rotation)

Click

Type

Listen

Games

Drucken