chapter 5-Inferential Results-Why did the authors reach their conclusion-What did they actually find? Flashcards
what are inferential results
• Inferential results are results intended to explain or predict a variable or variables
what do descriptive stat results allow us to know and explain. is this enough?
• Descriptive statistical results allow us to know and explain variables that we are interested in understanding but we have to go a step further to use that explanation to predict or infer how those variables may occur in the future. This can be done w inferential stats which are based ont he concepts of probability and statistical significance
in inferential stats how do you test for relationsips, differences, assoc among variables
by creating distributions of test stats that reflect variables having no connection between them, are unrelated, or are not differen
when do we say that a stat is statistically significant
• If a test stat falls out of the range of values that we would expect to occur 95% of the time, if there were no relationship among the variables
what is t value
- The t value is a test stat for differences in means between two groups.
- A t distribution hows how the t tests for hundreds of different samples of two variables that didn’t differ from each other in the real world would be distributed (see image on pg 93 it looks like a bell curve)
in this image the green section is where the t-test values will fall 95% of the time if the two variables tested aren’t different.t he red zones on either end of the normal curve are the areas where the t values will fall by chance alone 2.5% of the time if, in the real world the two variables tested aren’t different
what is p value
• P value represents probability
is defined as the percentage of the time the result found would have happened by chance alone.
So, if we infer from the finding of statistical significance that the difference, association, or relationship that we tested statistically is one that exists in the real world because we were unlikely to get our statistic by chance alone
are inferential methods used for Qnt Qal or both
• Inferential methods only used for quantitative
what is a confidence interval
does it usually include zero
• Confidence intervals state the range of actual values for the stat we are computing (eg difference in mean ages of students who do and do not choose acute care settings) in which 95 out of 100 values would fall
no doesnt include zero
how are CIntervals and p values expressed differentaly
• CIs are almost always stated for the 95% range, whereas the probability of getting the result reported if theres really no difference or relationship is gen reported as one of 3 possible percentages: 5% (p
what assumption is made p value vs CI
The relationship or difference tested is zero for p value
The relationship or ifference is that found in the data in CI
meaning P value vs CI
p value Gives the % of the time that wed get the test statistic by chance alon
CI gIves the range of values (biggest and smallest numbers) that would occur 95% of the time for the relationship or difference found
interpretation of p value vs CI
The smaller the p value the less likely that the test result occurred by chance alone
CI=The smaller the range, without zero in it, the more confident we can be that the test stat reflects the real world
is statistical significance an absolute guarantee that the values are really different or related in the real world.
no
Stat significanc means that theres less than 5% chance that the amount of relationship or difference found happened by chance
who do parametric and nonparametric studies refer to in gen
• These tests refer to the two broad classes of inferential ssts procedures that can be applied to numeric results from studies
what 2 criteria are required before you can apply a parametric stat
The numbers must gen be normally distributed (the freq distribution of the numbers is roughly bell-shaped and
The numbers must be interval or ratio numbers, eg age or intelligence scores (the numbers must have an order, and there must be equal distance between each value
what are nonparametric stats and some examples
• Nonparametric stats are used for numbers that don’t have a bell-shaped distribution and are categoric or ordinal (represent variables for which theres no est equal distance between each category, eg numbers used to represent gender or rating or preference for car color
wuld gender and litigation be para or nonparametric
what about age and education in yrs
• In predictors of postconcussive symp, gender and litigation would be nonparametric and age and education in yrs scores would be parametric
are t tests parametric or nonparametric
parametric
• Researchers use different types of stats to test for the same kind of relationship depending on the form of data collected..
.
for what 3 reasons do researchers use stat tests
o Look at differences between groups for on or more variables
o Look at relationships among two or more variables
o Look at relationships of factors within a variable itself
what are multivariate vs bivariate tests
bivariate used when looking at two variables or two different groups
from ppt: which kinds of tests Address Basic difference (Research question and hypothesis)
T-tests
One-Way Analysis of Variance (ANVOVA)
Repeated-Measures ANOVA
Pearson Chi-Square Test
Address Basic Association and Correlational research questions and hypothesis
Pearson Product-Moment Correlation Coefficient (r)
Spearman Rank-Order Correlation Coefficient
Simple and Multiple Linear Regression
ppt: which tests Address Basic Association and Correlational research questions and hypothesis
Pearson Product-Moment Correlation Coefficient (r)
Spearman Rank-Order Correlation Coefficient
Simple and Multiple Linear Regression
are these tests for independent groups multivaiate or bivariate results
ANOVA –parametric ANCOVA, MANOVA, one way ANOVA-parametric Kruskal-Wallis one-way ANOVA- nonparametric Repeated measures ANOVA- parametric Chi-square for independent samples
multivariate
are these groups used for related groups or independent groups and are they multi or bivariate
- T test (parametric)
- Sign test or median test (nonparametric)
- Mann-Whitney U (nonparametric)
- Wilcoxon rank test (nonparametric)
- Fisher exact test (nonparametric)
independent and bivariate
when looking at whether relationships between variables-have a natural connection between two or more variables what tests do you use
bivariate • Pearson r • Spearmon rho • Kendall tau • Contingency coefficient
multivaiate Multiple regression (parametric) Canonical correlation (parametric) Path analysis (parametric) Structural equation modeling (parametric) Discriminant analysis (parametric) Logistic regression (nonparametric)
if youre looking for a test for differences between two roups what to use
what is ti?
• T test computes a stat that reflect differences in the means of a variable for two different groups or at two different times for one group eg those who quit smoking and those who didn’t, men and women etc…in these examples one variable differentiates the groups. The variable can be anything that can be measured as a continuous number eg cost per pt visit, age
what does covary mean and when do you use it
• When two variables are connected in some way, they are said to be covary-2 variables covary when change in one are connected to consistent changes in the other eg height and wt covary in kids. Another covariance between # times doing procedure anddec of # of errors
correlation stat aka
correlation coefficient
2 important things about correlation statistic
o Is the number negative or pos. if both inc itll be pos (ht and wt example) in second example above the two variables move in opposite directions=negative relationship
o Note the magnitude of the number for a correlation coefficient. Because of how its calculated it can only have a range of values from-1 to +1. A perfect relationship between two variables-as one goes up the other goes up or down in same amount will have a value of either -1 or +1
is there generally perfect correlation in real life
does covariance tell us the cause of the connection
no
no
can you use correlations to predict
what kind of stats are these
egs of correlational stats (prob not imp)
• correlations are inferential stats that explain about relationships but cant be used to predict because they don’t tell us anything about which variable causes the other variable to change
- the most common type of correlation statistic is the Pearson product-moment correlation which uses r to represent that value of the bivariate relationship
- other examples of correlation stats: Spearman rho, Kendall tau, and the phi (the stat gives the strength of the covariance between wo variables
what kind of test would you use to look for differences among 3 or more groups
analysis of variance tests for differences in the means for three or more groups
what does ANOVA tell you and what is the letter that it uses and how is it used
- ANOVA compares how much members of a goup differe or vary from the members of other group
- The test stat in ANOVA is usually eh F ratio. Like other stat tests, the final test stat is then compared w a sset of stats one would get if there were no diff between groups
- The larger the F ratio the more variation between groups