Exam 2 Flashcards
What is the Null Hypothesis (Ho) and the Alternate Hypothesis
The null hypothesis, denoted by Ho, is usually the hypothesis that sample observations result purely from chance.
The alternative hypothesis, denoted by H1 or Ha, is the hypothesis that sample observations are influenced by some non-random cause.
Null Hypothesis: chance
Alternate hypothesis: not chance
What is a p-value
What does a p-value LESS than 0.05 mean?
What does a p-value GREATER than 0.05 mean?
P-value helps you determine the significance of your results. It is a number between 0 and 1. Closer to 0 means you REJECT the null hypothesis (in other words, it isn’t random chance, but some intervention influenced the result).
P-values less than 0.05 means it is significant that you can REJECT the null hypothesis. (0.05 means 5% chance or probability)
P-values greater than 0.05 means there is weak evidence against the null hypothesis so you fail to reject the null hypothesis.
In the majority of analyses, an alpha of 0.05 is used as the cutoff for significance. If the p-value is less than 0.05, we reject the null hypothesis that there’s no difference between the means, and so we conclude that a significant difference does exist.
If a measurement is less precise, how does that effect the association between 2 variables?
The less precise a measurement is, the lower the association is between two variables.
T or F: If a measurement is reliable and fairly precise, a strong association may be found.
True
*** What is the fundamental statistic for the analysis of INTERVAL and RATIO data to estimate the strength of the association between variables.
And what is it?
Regression Analysis
RAtio = RA = Regression Analysis = Pearson’s R
Regression analysis is used to understand which among the independent variables are related to the dependent variable, and to explore the forms of these relationships.
How does regression analysis look?
Graph with dots in relation together going up a line … graph relationship between independent and dependent variables. Or how the dependent variable changes when the independent variable is changed.
______ ______ regression is the simplest form of regression.
What is it?
Simple linear
Study and plot relationship between 2 variables (independent and dependent). X influences Y
What is pearson’s r or r value?
Designation of measuring the strength of the association between pairs of data that are interval or ratio
RA = Ratio = Regression Analysis = Pearson’s R
What is ANOVA
Analysis of variance, ANOVA: a statistical method for making comparisons between two or more MEANS; a statistical method that yields values that can be tested to determine whether a significant relation exists between variables.
What is a t-test
T-test: is simplest type/form of ANOVA … is an analysis of two populations means through the use of statistical examination
Difference between ANOVA and t-test
Example questions:
“Would you use a T-test or an ANOVA to compare 2 groups?”
“What about if you compare a group of patients w/ stroke vs. a group of patients w/ MS? Independent or dependent T-test?”
“What about if you were comparing a group of children, pre and post intervention? You are using the same group.”
A hypothesis test that is used to compare the means of TWO populations is called t-test.
A statistical technique that is used to compare the means of MORE THAN TWO populations is known as Analysis of Variance or ANOVA
T-Test: A simple ANOVA where you have 2 groups you are comparing. A control group against a select group with a condition.
Example questions:
- T-test
- Independent T-test
- Dependent T-test
What is multiple regression
Plot the relationship between several independent or predictor variables and a dependent or criterion variable.
X1 and X2 and X3 and X4 —–> all lead to Y
T or F: simple linear regression is r, and multiple regression is r^2
True
Multiple regression:
R2 values (0 to 1)
PPMC (Pearson’s r):
r values (-1 to 1)
On exam, she will have two variables with data, and you need to know whether they have low correlation or high correlation:
.00 to .25 or .00 to -0.25 .26 to .49 or –.26 to –.49 .50 to .69 or –.50 to –.69 .70 to .89 or –.70 to –.89 .90 to 1.0 or –.90 to –1.0
Little to no correlation Low correlation Moderate correlation High correlation Very high correlation
So if you see r = 0.62 and p = 0.03 … how do you interpret?
It is a moderate correlation, and less than 0.05 for p-value so we can reject null hypothesis since there is evidence that it is not chance.
*** T or F: Variables that are associated do not imply cause and effect.
Give example:
True! Just because there is an association between variables, doesn’t mean there is a cause and effect relationship. (Although measures of association cannot confirm cause and effect, these analyses can build useful predictive models).
Example: So does your socioeconomic status effect BMI? They may be associated, but they may not necessarily cause / influence each other.
The analysis used most often for ORDINAL DATA is a:
Spearman rank order correlation
Difference between Pearson’s and Spearmans
The Pearson correlation evaluates the linear relationship between two continuous variables. … The Spearman correlation coefficient is based on the ranked values for each variable rather than the raw data. Spearman correlation is often used to evaluate relationships involving ordinal variables.
T or F: Spearman r and Cramer’s V are nonparametric statistics
True
What is difference between parametric and nonparametric tests
Parametric:
- assumed NORMAL distributions
- interval and ratio data
- statistics based on assumptions about the population
- Have more statistical power
Nonparametric:
- Uneven distribution
- Ordinal, ranked, nominal data
- LESS powerful (use less info in calculation)
*** If your measurement scale is nominal or ordinal, would you use parametric or nonparametric?
If it is interval or ratio?
Nominal or Ordinal = Non-parametric statistics
Interval or ratio = Parametric statistics
** If you have this type of data below, list the type of test you would use and the statistical procedure:
- Interval or Ratio
- Ordinal
- Nominal
- Interval or Ratio: Pearsons r, Parametric
- Ordinal: Spearman r, Nonparametric
- Nominal: Cramers V, Nonparametric
SPEARMAN: rank ordinary gums
CRAMERS: nominal
Difference between sensitivity and specificity?
Which one has the condition?
Test SENSITIVITY is the ability of a test to correctly identify those with the disease (true positive rate). A
Test SPECIFICITY is the ability of the test to correctly identify those without the disease (true negative rate). D
Sensitivity = HAVING THE CONDITION. The number of people that are correctly diagnosed by exam.
Specificity = NOT HAVING THE CONDITION. The number of people correctly classified as not having the condition
Remember if you have the 4 box table, A is top left and is TRUE positive. D box is bottom right, and is TRUE negative. So boxes are …
ab
cd
What is A and D?
How do you calculate:
- Sensitivity
- Specificity
- PPV
- NPV
A = sensitivity (have condition) D = specificity (without condition)
- Sensitivity: a / a+c
- Specificity: d / d+b
- PPV: a / a+b
- NPV d / d+c
In other words:
ab a / a+b=PPV
cd d / d+c=NPV
SS
Sensitivity: a / a+c
Specificity: d / d+b
How do you calculate the +LR and -LR
+LR: sensitivity / (1-specificity)
-LR: (1-sensitivity) / specificity
What is SNout and SPin
SNout = Sensitivity … rules OUT a condition (if test is neg)
SPin = Specificity … rules IN a condition (if test is positive)
When sensitivity is High (close to 1), we can confidently RULE OUT the condition when the clinical test is negative.
When specificity is High (close to 1), we can confidently RULE IN the condition when the clinical test is Positive
Gold standard is what?
Xray or MRI
What are the 2 main criteria to know how VALID a test is
Sensitivity and Specificity
If you get a 0.88 for sensitivity and a 0.90 for specificity, is that a good test?
Yes
T or F: Knowledge of the LRs influences the level of certainty that a condition does or does not exist at the end of the examination.
True
A large +LR Rules IN or OUT a disorder?
A large –LR Rules IN or OUT a disorder
IN (so higher the positive number, more likely the disease/condition is there)
OUT (so more negative the number, more likely the disease/condition is not there)
Remember how to grade area under the ROC curve
.90-1.0 = Excellent .80-.90 = good .70-.80 = fair .60-.70 = poor .50-.60 = fail
What is cut off on ROC curve
Point where curve turns … a balance between sensitivity and specificity
What is STARD
STAndards for the Reporting of Diagnostic accuracy studies
4 steps of Bahr’s “Framework of Risk Identification and Injury Prevention”
1) Establish the extent of injury or severity (PROSPECTIVE STUDY)
2) Establish cause of sports injury
3) Introduce a preventative measure
4) Reassessing injury and improvement over time, and repeating step 1
Difference between Traumatic and Atraumatic:
Traumatic = contact or extreme accident Atraumatic = osteo bone decay over time that led to ACL injury
The Gold standard for studies of injury risk factors?
What is the other option?
PROSPECTIVE COHORT DESIGN … go forward in time
A large group of participants potentially “at risk” for injury of interest are baseline tested. Participants are followed over time to determine if they go on to suffer the injury.
Other option: Retrospective design (look back over time and compare group that had injury in past with control group going forward).
*** The prevalence of an injury or illness refers to the …
proportion of a sample that has a given injury or illness at a single time point. It is presented as a proportion or a percentage.
*** The incidence of an injury or illness …
… the number of new cases of the pathology in a given period of time.
Bathtub analogy related to prevelance, incidence, mortality, recurrance
Incidence is new amount of people coming in, prevelance is how many patients in study at time, mortality is how many die or leave.
Does rate have to do with TIME, or proportion?
RATE = number of cases OVER TIME Proportion = number of people (without time component)
(***) Know this … just know RATE has to do with TIME, and proportion does not.
*** EXAM QUESTION
(***) The simplest comparison between two measures of injury incidence is to
calculate the ratio of the injury incidence between two groups.
What is absolute risk reduction (ARR)
Absolute risk reduction (ARR): the difference between the injury rate/risk in the intervention group and the injury rate/risk in the control group.
What is numbers needed to treat (NNT)
(***) Numbers needed to treat (NNT): represents the number of patients that need to be treated with the experimental treatment to prevent one injury compared to receiving the control condition.
If your NNT is 1 patient Then that person is highly Likely to get prevented. Or It could take 3,000 patients Just to get one person to Have prevention.
How many patients do I Need to treat in order To get an effect in 1
If NNT was 5.56 it would mean 1 in 5 or 6 people. If NNT was 8.9 it Would be 1 in 9 people.
T or F: The closer to 1 the more relevant the study
True
4 things data is categorized as:
Nominal
Ordinal
Interval
Ratio
Nominal simply means “to name.” The assignment of numeric values for analysis of nominal data is arbitrary.
Ordinal data are ordered in a particular and meaningful manner (e.g., numeric pain scales)
Difference between parametric and nonparametric
Parametric statistics are appropriate for analyzing interval and ratio under most circumstances with normal distribution.
Nonparametric statistical methods of comparison are used to analyze nominal and ordinal data.
T or F: Variance and standard deviation are measures of dispersion for interval and ratio data.
True
T or F: Median and range value are reported for ordinal data and the mode for nominal data.
True
The result of an ANOVA is an F value.
T or F: The higher/larger the F value is, the better it is.
True
What is a MANOVA and when would you use it?
MANOVA refers to multivariant analysis of variance or cases where more than one dependent measure is analyzed simultaneously.
MANOVA is best applied when the investigator is interested in the effect of the independent variable(s) on the collection of dependent variables.
Just know this
You’d use a T-test with 2 groups. With 2+ groups, you’d use an Analysis of Variance (ANOVA).
ANOVA – figure out how much of a difference there is BETWEEN and WITHIN groups. The higher the number, the more you’ll REJECT the null hypothesis.
F (b, w) where b = variance Between groups, and w = variance Within groups. If it is Between groups, there is a significant effect. If the variance is within groups, there probably is NOT a significant effect.
The range of what a F value can be depends. If p value is significant, and your F value is high – you can reject the null hypothesis
Explain difference between Type I and Type II errors
Type I: False positive (you are pregnant to man)
Type II: False negative (you are NOT pregnant to lady)
Are wider or narrower confidence intervals (CI) better?
*** The wider the CI, the WORSE it is. The CLOSER they are, the BETTER **
Larger the CI, the more skepticism. The narrower, the more confident.
THE WIDER THE CI, THE WIDER THE VARIANCE
CI’s are reported at
So a wider CI means what?
95%
More variance
What is effects size
How is it measured
The larger the effect size the greatest difference between groups
Cohen’s D
0.2-0.49 represents a small effect.
0.5-0.79 represents a moderate effect.
≥ 0.8 represents a large effect.
MCID =
Minimally Clinically Important Difference
MCID is the smallest treatment effect that would result in change in patient management, given the side effects, costs, and inconveniences. It is also part of power calculations
What are t-tests
T- Tests are a special case of ANOVA in which there are only two sets of data in the comparison.
Are t-tests positive or negative?
Are F values positive or negative?
t values may be positive or negative. (F values are always positive.)
Why use a t-test vs. an ANOVA
The t-test compares the means between 2 samples and is simple to conduct, but if there is more than 2 conditions in an experiment an ANOVA is required. The fact the ANOVA can test more than one treatment is a major advantage over other statistical analysis such as the t-test, it opens up many testing capabilities
ANOVA … t test is simplest form of an ANOVA. Use a t test with 2 groups. If you compare two separate groups, do an INDEPENDENT t test. If you have same group, you do pre and post test, it is a DEPENDENT t test.
Gold standard for type of test:
Gold standard is MULTI-SITE LARGE SCALE RCT
What is a CONSORT
Consolidated Standards of Reporting Trials
Evidence-based, minimum set of recommendations for reporting randomized trials. It offers a standard way for authors to prepare reports of trial findings, facilitating their complete and transparent reporting, and aiding their critical appraisal and interpretation.
What is PEDro
What is a PEDro scale
PEDro = physiotherapy evidence database
A database with citations to RCT’s, reviews, and evidence based clinical practices to PT. There is a PEDro scale to help you determine the statistical significance of a certain study you seek out in that database.
Purpose is to help you identify which of the known or suspected RCTs are likely to be internally valid and has sufficient statistical information to make results interpretable.
The purpose of the PEDro scale is to help the users of the PEDro database rapidly identify which of the known or suspected randomised clinical trials (ie RCTs or CCTs) archived on the PEDro database are likely to be internally valid (criteria 2-9), and could have sufficient statistical information to make their results valid.
PEDro scale is that list of ?s we went over in class to ask about an article. The higher the score of “yes’s” they get, the more valid the study.
How to calculate prevelance:
a/a+b / c/c+d
Does parametric need to be normal distribution or unnormal
normal
Interval and ratio goes with _________
Nominal and ordinal goes with _______
Parametric
Nonparametric
Pearson r is for lineal regression, for multiple regression it is ________
R squared
4 steps of prognosis
????
Abstract should be how long?
150-250 words, or 3125 characters (including spaces)
My data is of a normal distribution, what kind of statistic is it?
Parametric
2 measures of association
Parametric and nonparametric statistics.
What is the likelihood ratio?
LRs are basically a ratio of the probability that a test result is correct to the probability that the test result is incorrect.
What is the fundamental statistic for the analysis of interval and ratio?
Regression Analysis
Most simple form of regression
Special case of regression
linear regression
ANOVA
A LR great than 1 =
A LR equal to 1 =
A LR less than 1 =
disease/condition is present
no change
disease/condition is not present (less likely)
Another name for simple linear regression
And it measures what?
PPMC = pearson product moment correlation
The strength of the association between pairs of data that are INTERVAL or RATIO
If you want to measure the association between multiple variables, use what?
Would you do a t-test or ANOVA if doing multiple variables?
Multiple regression
ANOVA (or MANOVA)
What is injury risk:
What is injury rate:
What is time loss injury:
Injury risk refers specifically to the probability of new injury per individual.
Injury rate specifically refers to the number of new injuries per unit of exposure time.
A time loss injury refers to an injury that forces a worker to miss work or an athlete to not participate in his or her sport.
What is relative risk
When assessing injury risks or rates between two groups, the method of comparison is the calculation of the relative risk of injury between the two groups.
RRR vs. RRI
ARR
RRR - Relative Risk Reduction: Negative
RRI - Relative Risk Increase: if experimental condition is found to lead to heightened risk of injury, the sign is positive not negative.
ARR - Absolute risk reduction: the difference between the injury rate/risk in the intervention group and the injury rate/risk in the control group. Control group incidence subtracted from intervention group incidence
What is the ideal NNT, and what is the worst?
1 … meaning that for every patient treated with the experimental treatment, an injury is prevented.
Worst is a million or anything high. Meaning I have to treat a million to prevent one injury.
NNT - Numbers needed to treat: represents the number of patients that need to be treated with the experimental treatment to prevent one injury compared to receiving the control condition.
If NNT was 5.56 it would mean
1 in 5 or 6 people. If NNT was 8.9 it
Would be 1 in 9 people.
Closer to 1 the better. 1 in 15 is
Not really an effective treatment.
1 in 2 is very effective
What method would you use to know these:
What is the reduction in risk of falls with group cycling compared to a control?
How many patients with knee OA do I need to treat within a community cycling program to prevent 1 patient from having a fall?
What is the reduction in risk of falls with group cycling compared to a control? (ARR)
How many patients with knee OA do I need to treat within a community cycling program to prevent 1 patient from having a fall? (NNT)
What does this mean:
Twelve separate 2x2 mixed-model analyses of variance (ANOVA’s) were used to analyze EMG activity.
** They did 12 separate ANOVA’s on 2 groups, with 2 testing points (pre and post)
Explain Type I and Type II errors
Type I error occurs when a null hypothesis is rejected when in fact the null hypothesis is true / population differences do not exist. The alpha value is really the level of risk of Type I error.
You reject null when null is true.
Type II error occurs when a null is not rejected when null is false / yet a study of the population would reveal differences between groups.
You don’t reject null, when null is false.
If your null hypothesis is true, you get what error?
If your null hypothesis is false, you get what error?
Type 1 error (FALSE POSITIVE)
Type 2 error (FALSE NEGATIVE)
When a type II error is made:
________ was not adequate
So to decrease risk of Type II error, what is required?
power
or how powerful the experiment
Statistical power
How can researchers impact power?
Sample size
Does statistical significance mean there is clinical significance?
No
CI’s are a way to represent how good an ______ is
estimate
What does MCID mean?
Minimally Clinically Important Difference (between groups)
t values of t-tests are positive or negative?
F values are positive or negative
can be positive or negative
always positive
What is method for calculating effect size
Cohen’s d
T or F: The larger the effect size the greatest difference between groups
True
0.2-0.49 represents a small effect.
0.5-0.79 represents a moderate effect.
≥ 0.8 represents a large effect.
So 0.9 is a LARGE effect. 0.1 is small effect.
T or F: statistics do not prove anything
true
T or F: Do not read to accept the conclusions of a research report as an absolute or final answer.
True
How do you estimate MCID
ROC curves
Know this
(**SOS **) MCID is the smallest treatment effect that would result in change in patient management, given the side effects, costs, and inconveniences. It is also part of power calculations
Is there more statistical significance with t-tests or ANOVA’s?
ANOVA’s
With levels of evidence for treatment outcomes, what is gold standard
The “gold standard” is Systematic Reviews with Meta-Analysis from well designed multi-site RCT’s
T or F: If a trial is sponsored by a company or group, you can expect a little bias by the researchers.
True
Know
Before a study is conducted, the researchers should perform a power analysis to determine how many subjects are needed to test the hypothesis and minimize the potential for errors in hypothesis testing (Type 1 and Type II).