Survey & Questionnaire Analysis Flashcards
Descriptive correlation
Describes characteristics of a sample: Central tendency (mean, median, mode) Dispersion (kurtosis, variance, standard deviation) Symmetry (skew)
What are the two methods of correlation?
Descriptive
Inferential
Inferential correlation
Used to make inferences from a sample to a pop
Normal distribution is
The mean = median = mode (bell curve)
Variance is calculated by
Take each score and subtract the mean
Total deviation is calculated by
Adding all the variances. Squaring this becomes the sum of squared errors
Average of the error (variance)
Sum of square error divided by the n -1
Alternative hypothesis means
An effect does exist
When talking about hypothesis we don’t say it is proven or correct we say it is
Supported or unsupported
If we see a certain result/quality in our research we either ……… Or ……… The hypothesis
Reject or retain
Standard error tells you if
2 means actually come from the sand population
Alpha is used to decide if
A result is significant or not
Alpha represents an
Error rate. How much you’re willing to be wrong
Alpha rate is
.05
More conservative alpha rate is
.01
1st question Inferential stats addresses
Relationship between variables (correlation, regression)
2nd q inferential stats addresses
Difference between groups on variable of interest (T-test, ANOVA, ANCOVA)
3rd question inferential stats addresses
Predicting group membership (discriminant function analysis)
4th question inferential stats addresses
Underlying structure (factor analysis)
3 types of correlation relationship
Linear (positive or neg)
Curvilinear (relationship does not fall on a straight line)
No relationship
Spurious correlations
Bed sheet tangled death and ski field accidents
Use scatter plots to look at
Correlation/relationships
Positive relationship in correlation means that as
As one variable goes up so does the other (ie grades and editing time) same direction
Negative relationship
As one variable goes up the other goes down
When looking at scatter plots/grams you need to see if it:
Straight line Monotonic (has one line) Relationship type Cluster Gaps Outliers
Covariance is
Are two variable associated. Eg if changes with one variable result in changes in the other.
How do we calculate covariance?
Calculate total amount of deviation
Cross product deviations
Used to examine covariance. Take deviations of both variables and multiply them. If both variances positive then it’s positive. If one of each then it’s negative.
Covariance is NOT a
Standardised measure
To convert covariance into a standardised measure we convert it into….
Pearson’s r
Pearson’s converts our scores into
Z scores
r ranges from
-.1 - -.5
-.1 is a
Small effect
-.3 is a
Medium effect
-.5 is a
Large effect
r is so useful as it shows us
The shared variance that 2 variables have. How much overlap. This is done by squaring the r
Many different types of correlation tests. How to decide which one to use?
Depends on your measurement
We use pearons r for measurements when they are
Interval or ratio
Pearson’s r will only catch a
Linear correlation
Pearson’s r is a ……….. Statistic
Descriptive
If sample size is large enough it becomes very easy to get
Significance
Psychometrics uses two criteria to address quality
Reliability
Validity
Reliability is
Concerned with internal properties of a scale. Reliability assessed using correlation. Weighing the gold…how reliable is the scale to CONSISTENTLY produce the same results? Consistency or stability of a measure of behaviour
Validity is
Concerned with external properties of a scale. Does it correlate with other variables/scales? Does the scale measure the construct that you think it is measuring?
It has to be reliable to be…
Valid
Doesn’t have to be valid to be
Reliable
Any measure of behaviour has 2 components…
True score (real score on variable) Measurement error
Confidence interval is
How sure we are that a true score falls between a certain range
4 types of reliability
Internal consistency
Test-retest
Inter-rater
Parallel forms
5 methods used to test internal consistency
Split-half Inter-item Item-total Alpha reliability Item analysis
Split-half reliability takes your items and
Randomly Splits items into 2 to if there is a good correlation btwn to halves. Good correlation = good reliability
Alpha reliability splits items in half to see if they correlate then
Randomly splits them again. Then again. Does every possible split and averages an overall number
Alpha is a
Coefficient
Alpha ranges from
-1 to +1
An alpha .7 and above indicates
Acceptable internal reliability for a set of items
.6 alpha can be argued to be acceptable if
Your study is based on introspection/ qualitative measures
If you have a - (negative) alpha
Something has gone wrong
Inter-item reliability is
When you take all of the correlations for each of the pairing of items and get the average
Item analysis is
Internally and reliably Differentiates between varying attitudes if ppl doing your questionnaire. Is there a difference btwn high raters and low raters
Test re-test reliability
Give test and then later after some time give same test again.
Test-retest measures
Consistency
Test-retest shows a correlation when scores are less than
.7
String correlations btwn scores in Test-retest indicates that the construct being measured is
Stable
Inter-rater reliability is
Agreement btwn 2 raters on scores they have given on a measure of behaviour
In inter-rater reliability the correlation of ……is considered acceptable
.8
Parallel forms of reliability is
When there are two versions of the same test given on different occasions
