Lecture 17: Validity And Assumptions Flashcards
Internal validity
How much the study design makes sense
External validity
How applicable they are to real life
Construct validity
How well the operationalization captures the construct
Demand characteristics
Pressures on the subjects to behave a certain way that isn’t how they usually would behave in the real world
Statistical validity
How appropriate the statistical analysis for the types of variables being studied/research quests asked are
Statistical assumptions
Things that must be true for our analysis to be valid
(p-value is uniform when n0 is true, 95% CI to actually have a 95% rate of containing the true value of the parameter))
In statistics calling something an “assumption” doesn’t mean we can:
Assume it’s true. It means if it isnt true out analysis may be false.
All statistical tests we’ve talked about assume _______ _______
Independent observations
Independent observations
- Each data point should be independent of each other data point eg. No subjects score influences another score, no subject is related to another subject like siblings
Parametric tests and name an example
Require certain assumptions about the population distribution eg. Pearson correlation tests
Unpaired t-tests assume the sampling distributions of the two means are _______, and paired t-tests assume the sampling difference of the mean difference is ______. This is always true of the variables themselves are _______ ________ in the population. It is also approximately true if the sample size is ______ (Due to a certain theorem)
NORMAL! And NORMAL!
Normally distributed
LARGE
Central limit theorum
Using the poorer standard deviation for an unpaired t-test assumes the two population variances are ___
EQUAL!
Pearson correlation tests assume ______ _____
Bivariate normality
Bivariate normality means 3 things (3)
- Both variables are NORMALLY DISTRIBUTED in the population
- The relationship if any between the variables is LINEAR
- The relationship between the variables is HOMOSCEDASTIC
