Final: Lecutres 22-24 Flashcards
Nominal (Dichotomous/Binary)
- No order or Rank, Non-Ranked Categories
- No magnitude/ no consistency of scale/ no rational zero
- Dichotomous breaks all the rules! Call it nominal b/c it’s only two. (Ex. high blood pressure vs. normal)
- Gender and Handed-ness ex.
3 Key Attributes of Data (variables)
- Magnitude (or Dimensionality) bigger is more, lower is less
- Consistency of scale (or Fixed Interval) equal, measurable spacing between units.
- Rational Zero (ex. blood pressure can not be negative)
•Each attribute can be assessed with a “Yes” or “No” response
Ordinal (Ranked Categories!!)
- Non-Equal-Distance
- Yes magnitude/ No consistency of scale/ No rational zero
- Levels of Intimidations, pain scale ex.
Interval/Ratio (Order/Magnitude & Equal intervals-of-scale)
- Yes magnitude/ Yes consistency of scale/ No (Interval) or Yes (Ratio) Rational zero
- All numerical scales with TRUE units
- Number of living siblings and Age Ex.
T/F After data is collected, you can always go up in specificity/detail of data measurement (levels).
•FALSE, you can go down, but never up!
Is a mental health scale, where they rank certain questions with strongly agree/disagree ect. interval or ordinal?
- Just for each individual question, it’s Ordinal.
* But if you rank each question and add up the scores, it’s Interval
What data level is each survey item? (A-Nominal, B-Ordinal, C-Interval)
- Age
- Sex
- Current Occupation
- Months Homeless
- Regarding your overall stress level the last 3 months, how often have you felt: out of control? sick? ect.
- How would you describe your overall health? very good, good, ect
- How many times in the last year have you seen doctor/dentist?
- C
- A
- A
- C
- B
- B
- C
Measures of Central Tendency and Dispersion
- Mean/Median/Mode
- Outliers
- Minimum/Maximum/Range
- Interquartile Range (IQR)
Variance
•Difference in each individual measurement value and the groups’ mean
Standard Deviation (SD)
•Square root of variance value (restores units of mean)
Parametric tests
- Stats test useful for Normally-Distributed data
* Symmetrical plot
Positively Skewed Plot
- Asymmetrical distribution with one “tail” longer than another
- Skewed anytime the median differs from the mean!
- When mean higher than median, skewed right (positive)
In a Negatively Skewed graph, the mean is _______ than the median.
•Lower
Required Assumptions of Interval Data
- Normally-distributed around a known mean
- Equal variances (SD) Levene’s test** assess for equal variance between groups
- Randomly-derived and independent
Handling Interval data that is not normally-distributed
- ALWAYS run descriptive statistics and graphs
- Use a statistical test that does not require the data to be normally-distributed (non-parametric tests)
- Transform the data to a standardized “score” (z-score); hoping that this transformation will cause the data to become normally-distributed in order to use a parametric test
Type __ errors rejects the Null hypothesis when you shouldn’t, and Type __ is when you should reject it, but the data says not to.
- 1, (false positive)
* 2 (false negative)
Statistical tests compare differences in variables or to evaluate relationships between them:
- A test statistic value is calculated, then,
- Compared to the appropriate table of probabilities for that test, then
- A probability (p)** value is obtained, based on the probability of observing, due to chance alone, a test is statistic value as extreme or more extreme than actually observes (probability of making type 1 error)
If p value is lower than pre-selecting a priori value (usually .05) then we say it’s?
- Statistically Significant
* If less than .05, we REJECT the Null Hypothesis (just says there is a difference, not by how much)
Interpretation of p value***
•Be able to say what p value represents on test (EXACT WORDS) 5 different ways:
- The probability of making a Type 1 error if the Null Hypothesis is rejected.
- The probability of erroneously claiming a difference between groups when one does not really exist
Impacts to Statistical Significance
- Power: the ability of a study design and its methodology to detect a true difference if one truly exists (The level of accuracy in correctly accepting/rejecting the Null Hypothesis)
- Sample size: the larger the sample size, the greater the likelihood of detecting a difference if one truly exists. (Increase in Power)
Sample Size Determination
- Difference between groups deemed significant: the smaller the difference between groups necessary to be considered “significant” the greater the number needed
- Baseline rate of outcome (known/estimated)
- Alpha and Beta Error rates (power) **Add in anticipated drop-outs or loss to follow-ups
Statistical significane
- Comparisons of groups generates only a statistical estimate of the “true” yet unknown difference between groups (a Point estimate)
- Spread (V/SD) in estimates of group comparisions can aid in interpretation
- Confidence Interval (CI) level of confidence you believe reality (the real difference) is located
*If CI crosses __ (for OR/RR/HR) or __ (interval variables) = NOT SIGNFIICANT
- 1.0
* 0.0
4 Key Questions to Selecting the Correct Statistical Test
- What TYPE OF DATA is being collected/evaluated? (does it have MAGNITUDE? does the the data have a fixed, measurable INTERVAL?)
- What TYPE OF COMPARISON/ASSESSMENT is desired? (correlation test)
- HOW MANY GROUPS are being compared? (2 or 3 or more)
- Is the data INDEPENDENT or RELATED (PAIRED)? (data from the same (paired) or different groups (independent) (related data come from same human, before/after, pre/post ect)
Correlation provides quantitative measure of the _______ and _______ of a relationship between variables.
- Strength
* Direction
Nominal Correlation test =
Ordinal Correlation test =
Interval Correlation test =
- Contingency Coefficient
- Spearman Correlation
- Pearson Correlation (p>0.05 just means there is no Linear correlation, there still may be a Non-linear correlation present)
Time-to-Event/Event-Occurrence—> survival test
- Nominal Survival test= Log-Rank Test
- Ordinal: Cox-Proportional Hazards test
- Interval: Kaplan-Meier test
- All can be represented on Kaplan-Meier curve
Regressions
- Provide a measure of relationship between variables by allowing the prediction about the dependent, or outcome, variable (DV) knowing the value/rank of others independent variable (IV)
- Also able to calculate OR for a Measure of Ass.
Know sheets to look at the rest of the tests
•Look at last slide of class 22-24 to practice!