Karteikarten
General MCQ Statistics 3
Karten in diesem Stapel (32)
- 1what type of data is a single proportion?categorical
- 2what is a single proportion?single sample of 'n' individuals that either possess or do not posses a characteristic
eg. pregnant or not pregnant - 3if we toss a coin and get 4 heads, what is the proportion of heads?p = 4/6 = 0.67
- 4proportion example of 64 women with bleeding disorders, 27 reported that they experience bleeding gums once a week what is our confidence interval for the prevalence of bleeding gums in the population of women with bleeding disordersp = 27/64 = 0.422 (42.2%)
STANDARD ERROR OF PROPORTION = square root of p(1-p)/n
=0.0617
CI for the proportion = +/- (1.96 x 0.0617 (SE))
=0.121 (12%)
Proportion is 42.2% with 95% confidence intervals (30.1% to 54.3%) - 5standard error of proportion equation?square root of (p(1-p)/n)
- 6what is another word for proportion data?binomial data
- 7what does correlation measure?measures the degrees of association between to 2 paired variables
- 8how is the raw correlative data plotted?scatterplot
- 9if there appears to be a linear correlative relationship, how can you then test whether it is significant or not?pearson correlation coefficient (r)
- 10how does the pearson correlation coefficient calculate r?measures how close the data points are to a straight line that best describes the linear relationship
- 11PCC - r = 1? r = 0? r = -1?+1 = perfect straight line with positive slope
-1 = perfect straight line with negative slope
0 = no linear relationship between x and y - 12what are the r effect sizes?0.1 = small effect
0.3 = medium effect
0.5 = large effect
ALL PLUS OR MINUS - 13when is correlation r misleading?when there is a non-linear relationship (r)
outliers
subgroups of data
eg. a curve rather than a straight line - 14what should you do if you are assessing correlation but there is a far off outlier?if it heavily influences your results, remove data point and re-run analysis
if it changes a lot then should report both values with and without the outlier - 15if the correlation line is curved, should you use Pearson's Correlation Coefficient?no
- 16why does correlation not mean causation?could be a 3rd variable
- 17with regard to correlation, what does r =?size of association/ effect size
- 18what does rsquared = ?COEFFICIENT OF DETERMINATION
the proportion of the total variance that is shared by the 2 variables - 19how can you calculate 95% confidence intervals for r squared?square the 95% confidence limits of r
- 20what are the 6 assumptions of Pearson Correlation Coefficient?1) continuous data on an interval or ratio scale
2) data are sampled at random
3) x and y are independent observations (measure independently)
4) x values are measured independently and not controlled
5) both x and y values must be sampled from a single population that is normally distributed
6) linear relationship between x and y - 21if the assumptions for PCC are not met, what is the non-parametric equivalent?Spearman's Rank Correlation
calculates rs that you can't square
relationship doesn't have to be linear - 22with correlation, can we conclude causality or directionality?no
- 23if rsquared = 0.11 then how much the variation can we attribute to the two measured variables?11%
- 24what does simple linear regression tell us?how well the data fits a theoretical model
used to:
fit curves to data
look for trends in data
predict values of y based on values of x
adjust for effects of confounding variables - 25main use of regression?to fit a straight line to data and use it to predict the values of y based on x
- 26change in independent variable causes a change in dependent variable, can this occur the other way round?no, regression is unidirectional
- 27in what direction are the residuals taken from? x axis? y axis? both?residuals only taken in the y-direction
- 28how does regression give us a predictive relationship? equation of line?y = a + bx
plug values into this and we can predict values - 29when looking up the p value using the calculated t for regression, you must use 2 degrees of freedom, why?2 because one for the mean of x and one for the mean of y
- 30what are the simple linear regression assumptions?1) x and y are asymmetrical - x predicts y (relationship MUST be unidirectional, shouldn't work the other way round)
2) linear/ straight line relationship between x and y
3) the independent variable (x) is measured without significant error (all the variability in the data is present in the y variable) - AS SHOWN BY THE RESIDUALS ON THE GRAPH
4) for each x value - the y value is sampled from an approximately normally distributed population of y values
5) homoscadedascity - the standard deviation of all y values is constant for all x values - 31what are the 3 questions you can ask yourself to decide whether simple linear regression is the correct choice of test?does the scatter appear to be normally distributed
is the average distance from 0 similar for all values of x
is there a relatively even scattering of points without clumping
IF THE ANSWER IS NO TO ANY OF THESE QUESTIONS, linear regression would not be appropriate - 32CONTINGENCY TABLES
Informationen
- Beschreibung
- 2nd Year Statistics Karteikarten am General MCQ Statistics 3, erstellt von William Chapman am 11/05/2016.
- Erstellt von
- William Chapman
- Aufrufe
- 9
- Bewertung