Lesson 5 (Stats) Flashcards
Nominal Scale
catagorical, distinct categories, least complex, fewest mathmatical operations
Ex: ice cream flavors, political parties, dietary restrictions
Ordinal Scale
assigned values to data based on rank or order, second least precise scale, not proportionally spaced
Ex: Army rankings, how depressed are you?
Interval Scale
involves the use of numbers wiht equal units of measuremnts ,
ex: fahrenheit, has no true zero temp or fixed beginning, SAT scores
cannot calculate ratios between scores
Ratio Scale
equal number of units with a true zero,
weight, or Kelvin temp scale
most precise of all the scales
Descriptive Statistics
data that is representattive of a sample of the population
typically refers to reporting means, standard deviations, falls on a normal curve
Inferential Statistics
generalizing data from a sample back to a population using probabilities and hypothesis testing to make inferences about the population from a sample
Assumption of Normality
extent to which a distribution of scores approximates the standard normal curve
Assumption of Linearity
extent to which 2 variables correlate in a linear fasion, some relationships may be more curved or circular, and violate this assumption, requiring a nonparametric statistic method ot analyze the data
Assumption of Independence
scores must be independent of each other, or influencing each other in any way, Ex: pre-test post test affects each other
Assumption of Homogeneity
data differs from each other similarly Ex: F Maximum test or Levine’ test
Central Tendency
math that is used in descriptive stats
Mean
the average of all numbers: Add all numbers and divide by the number of numbers
Mode
most frequent
Range
difference between the highest and lowest scores
Median
middle point, 50% of scores above and 50% below
Calculated by adding 1 to the total number of observations and dividing by 2, if a researcher has 11 scores, the median is the 6th score when arranged from high to low, if you have an even number, take the 2 middle scores and add them together and divide by 2
Bimodal distribution
set of data that has 2 modes, important to find out why this is happening if there is a distance between the modes
homogenous distribution of scores
a distribution laking variability
varience
how close scores are to the mean
standard deviation
square root of the variance
standard bell curve distribution
this illustrates the way the scores vary around the mean by standard units
left skew distribution
Or Negative skew, mode above median (the tail is on the left) mean is to the left of the median
right skew distribution
Or positive skew, mode to the left of median, mean is to the Right of the median, tail to the right
quartile distribution
Instead of the standard deviation percentages, it goes by 25% sections
stanine distribution
Break the bell curve into 9 different parts
Z-Score
Same as Standard deviation, with a population raw data more than 30
Range from -3 to 3
T-Score
Standardization from a population of less than 30 people
Range from 20-80
Levels of Significance
The chance of an error occurring in the rejection of the null hypothesis, aka type 1 error
Alpha Value
Aka P-Value: usually .5
Levels of significance which is the probability of accepting or rejecting a null hypothesis
Null Hypothesis
Refers to a conclusion that there are no differences between compared groups
Type I error
Or alpha error, occurs when after analyzing the data, there is a decision to reject the null hypothesis, when it is actually true (such as in the case of a confounding variable)
Type II error
Or beta error, occurs when the researcher accepts the null hypothesis, although there is in actuality, a difference
One tailed test
Directional experimental hypothesis, predicts the outcome
Two tailed test
Non-directional experimental hypothesis, there will be some effect, open ended
Parametric statistics
Data must be interval or ratio, and meets the 4 basic assumptions
Non-parametric statistics
Used with nominal or ordinal types of data, or the basic assumptions are not met
Chi-squared
Only one able to analyze nominal data
Parametric and non-parametric test (but not as strong as other parametric tests)
Can see what we expected to see vs what was observed, calculated by summing the total number of responses and divide by two
Wilcoxon signed rank test
Statistical procedure used to compare differences btwn paired ranks or ordinal data
Mann-Whitney test statistic
Statistical procedure used to compare difference btwn groups when data violates one plus assumption’s underlying inferential stats (normality, homogeneity etc)
Ordinal data, skewed data, matched pairs
Kruskal-Wallis
Randomly sampled, test for 3 groups of IVs
T-test
Used to compare the differences btwn groups and within
The difference btwn groups must be greater than the differences within the group
Two groups (control and intervention groups)
ANOVA
Stands for Analysis of Variance (aka F test)
Used to compare the differences within each group, with differences btwn 2+ groups
Usu used when there is one IV and typically 3+ groups
DV must be in interval or ratio scale
MANOVA
ANOVA with MULTIPLE analyses, EX: one group studies 20 min, one group studies an hour, + control group, analyze the data for math AND history classes = multiple analyses
Two-way ANOVA
Use two IVs to study one DV (or outcome)
Ex: factors or IVs (3 levels of exercise, including one controlled group) plus two groups for calorie intake (one control, one low calorie) and you see how both factors affect weight loss
MANCOVA
One IV, plus a control variable, looking at 2 DVs (outcomes)
Positive correlation
Btwn 0 and +1
When this happens this other thing also happens (or mostly)
Relationship btwn this thing and that thing happens
Negative correlation
A increases, B decreases (usually)
Values btwn 0 and -1
Closer to 0 is weaker correlation for positive or negative, closer to +1/-1 is strong
Below +/-0.3 = not significant
Pearson - R
Most frequently used correlation stat
Only used with integral or ratio data
Only used for single correlations
Spearman Rho
Ranked order correlation
Used to determine correlation btwn 2 variables when one of the variables is based on rank order data
Pearson r is much stronger
Structural Equation Modeling
Or SEM, a type of regression (which is a thread to internal validity)
Help predict the range the score falls in on the retest
Predicts the very high and very low scores will move towards the mean
Multiple regression
A statistical procedure used to calculate the relationship btwn a predicted variable and several predictor variables
Classic example: factors that predict success in college
Factor analysis
Technique that is used to reduce a large number of variables into a fewer number of factors
ANCOVA
An ANOVA, but with a covariate, a control variable
Two way MANVOVA
Two sets of IVs, two DVs (outcomes)