QUANTITATIVE ANALYSIS CHEAT SHEET

G RADUATE R ESOURCE C ENTER , U NIVERSITY OF N EW M EXICO

1 UNDERSTANDING DATA 2 D ESCRIPTIVE S TATISTICS 3 I NFERENTIAL S TATISTICS 4 I MPORTANT I SSUES

Useful Definitions and Terms Describe Variables Definitions and Terms Data is Messy
Types of studies: Frequency Tables: Inferential Statistics: To find relati- Data Cleaning: Correcting or removing inaccu-
Observational - Data without an intervention Absolute - Count/tally of occurrence of each onship/association between two or more groups rate or non-uniform entries from the data set
e.g, cohort study individual category Inspect: Detect unexpected, incorrect, and
Experimental - Data with an intervention Relative - Percentage of each category relative to Parameter: Unknown population characte- inconsistent data
e.g, Randomized controlled trail the total ristics such as population mean (µ) mean and E.g., Check for outliers, duplicates
Cumulative - Summation of relative frequencies variance (σ 2 )
Population: Entire collection of individuals or Cross tabulation - Grouped absolute frequency Clean: Fix or remove the anomalies
objects about which information is desired. table for two or more variables Statistical hypothesis: Statement about the discovered
parameters of one or more population. E.g., Transpose data structure, delete
Sample: A subset of the population that is Central Tendency: Null hypothesis (H0 ) - Usually, no relationship duplicates and inconsistent values
selected for analysis. Mode - The most commonly occurring value between two variables being tested
Median - Value that lies in the middle Alternate hypothesis (Ha ) - Usually, two or About Assumptions
Random sampling: Every possible sample Mean - The mathematical average more variables are somehow related All of the tests or models have certain assumpti-
of a certain size has the same chance of being ons
selected. Measure of Dispersion: Statistically Significance: Result of statisti-
Range - Highest value minus lowest value cal test being performed - the relationship is not All models assume that the observations
Types of variable: Standard Deviation - how much each individual due to random chance are independent and are randomly sampled
Qualitative - the information is non-numeric value is dispersed around the mean from the entire population of interest
Quantitative - the information is numeric Variance - Standard deviation squared α level or p-value: The probability or per-
centage of rejecting the null when it should not The relationship between two variable de-
Variable Measurement Methods and Formulas have been rejected pends on the method used to fit the data
Quantitative variables: Mean:
Discrete - Counts of individual items or values Sum of all values in the sample Independent variable: The presumed cause
M ean (ā) = Reporting and Discussing Results
Continuous - Measurements of continuous N umber of values in the sample in an analysis Figures:
or non-finite values Graphs of a variable - E.g., histogram, bar graph
n
1 X a1 + a2 + · · · + an Dependent variable: The presumed effect in Graphs of two variables - E.g., scatter plot
Qualitative variables: ā = ai = an analysis
n i=1 n
Binary/Dichotomous - Occurs only in groups Tables:
Nominal - Groups with no rank or order Comparing Between Two or More Groups Descriptive - Usually reports mean, standard
Ordinal - Groups with rank or order Standard Deviation: t-test: deviation, sample size, or frequency
v Paired t-test: Hypothesis testing - Report α level or p-value,
n
Levels of Measuring Variables Analysis of Variance (ANOVA):
u
u 1
group mean, sample size
X
s=t (ai − ā)2
Nominal: Groups with no rank or order (e.g., n − 1 i=1 Regression - Report regression coefficient,
academic department- Economics, Math, Biology) Association and Correlation p-value, sample size
Chi Square
Ordinal: Groups with rank or order (e.g., Pearson
letter grade- A, B, C) I NFERENCE FOR THE M EAN N EED H ELP ?
Explaining a Dependent Variable
Interval: Has order, distance between rank Useful Concepts Simple vs. Multiple Regressions Contact Us
is measurable, lacks natural zero (e.g., GRE Confidence Interval: Interval where we believe Linear vs. Non-linear Regressions Graduate Resource Center
scores) with C% confidence the population mean lies. E.g., - ordinary least square, logit/probit, multinomial Mesa Vista Hall, Suite 1057
Typically C = 95%, 99%, 90% logit Phone: 505-277-1407
Ratio: Has order, distance between rank is Email: [email protected]
measurable, has natural zero (e.g., height) Margins of Error: How many percentage Website: https://unmgrc.unm.edu/
points will differ from the real population value