K now About S tatistics in 45 Terrifying M inutes Kathy Herbst USA Goedele Beckers Netherlands Martin Kaefer USA Magdalena Fossum Sweden ESPU Research Committee Educational Session ID: 1047857
Download Presentation The PPT/PDF document "Everything You S hould" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. Everything You Should Know About Statistics in 45 Terrifying MinutesKathy Herbst (USA), Goedele Beckers (Netherlands), Martin Kaefer (USA), Magdalena Fossum (Sweden)ESPU Research Committee Educational Session
2. Data Distribution
3. Data DistributionAllows you to understand your cohortAllows you to determine how well the true mean is represented in your dataIt determines which statistical tests should be usedAssists with data interpretation
4. Data Distribution
5. Data Distribution68% within 1 SD of the mean95% within 2 SD of the mean99.7% within 3 SD of the mean68-95-99.7 rule (3-sigma rule)
6. FrequencyAge at Surgery (Months)Mode MedianMeanData DistributionDistribution of Age at Surgery Among Patients Undergoing Ureteral Reimplantation
7. FrequencyAge at Surgery (Months)IQRMedianRangeData DistributionDistribution of Age at Surgery Among Patients Undergoing Ureteral Reimplantation
8. Bivariate Analysis: The simultaneous analysis of two groups or variables.Bivariate Analysis explores the concept of relationship between two groups/variables:AssociationThe strength of this associationDifferences between two variablesSignificance of these differences.Comparative Statistics
9. CharacteristicTotal VUR ProceduresReimplantationInjectionp-ValueCohort14,4307,045 (49)7,385 (51)–Age at Initial Intervention (yrs)4.7 (2.5 – 7.2)4.2 (2.1 – 6.7)5.2 (2.9 – 7.7)0.001aFemale11,999 (83)5,605 (80)6,394 (87)–Age at Intervention (yrs)4.9 (2.8 – 7.3)4.6 (2.4 – 6.7)5.3 (3.2 – 7.8< 0.001aMale2,431 (17)1,440 (20)991 (13)–Age at Intervention (yrs)3.2 (1.5 – 6.6)2.7 (1.3 – 5.6)4.0 (1.8 – 7.8)–Data in table are given as n (%) or median (25th, 75th percentile)a Mann-Whitney U testHerbst K, Corbett ST, Lendvay TS, Caldamone AA. Recent Trends in the Surgical Management of Primary Vesicoureteral Reflux in the Era of Dextranomer/Hyaluronic Acid. J Urol. 2014 May:191(5):1628-1633.Patient Characteristics Among Patients Who Underwent VUR InterventionBivariate Comparison
10. CharacteristicTotal VUR ProceduresReimplantationInjectionp-ValueCohort14,4307,045 (49)7,385 (51)–Age at Initial Intervention (yrs)4.7 (2.5 – 7.2)4.2 (2.1 – 6.7)5.2 (2.9 – 7.7)0.001aFemale11,999 (83)5,605 (80)6,394 (87)–Age at Intervention (yrs)4.9 (2.8 – 7.3)4.6 (2.4 – 6.7)5.3 (3.2 – 7.8< 0.001aMale2,431 (17)1,440 (20)991 (13)–Age at Intervention (yrs)3.2 (1.5 – 6.6)2.7 (1.3 – 5.6)4.0 (1.8 – 7.8)–Data in table are given as n (%) or median (25th, 75th percentile)a Mann-Whitney U testHerbst K, Corbett ST, Lendvay TS, Caldamone AA. Recent Trends in the Surgical Management of Primary Vesicoureteral Reflux in the Era of Dextranomer/Hyaluronic Acid. J Urol. 2014 May:191(5):1628-1633.Patient Characteristics Among Patients Who Underwent VUR InterventionBivariate Comparison
11. Categorical (nominal) Data: Has two or more categories, but order doesn’t matter.Continuous Data: A variable that can take on any value between its minimum value and its maximum value.Bivariate Comparison
12. T-test: Assesses whether the means of two groups are statistically different from each other.Mann-Whitney U test: Determines if the proportions of one variable are different depending on the value of the other variable.Small Sample SizesDoes not require assumption of normal distribution.Bivariate Continuous Comparison
13. T-test: Assesses whether the means of two groups are statistically different from each other.Mann-Whitney U test: Determines if the proportions of one variable are different depending on the value of the other variable.Small Sample SizesDoes not require assumption of normal distribution.Bivariate Continuous ComparisonNormal Distribution = Parametric testsSkewed Distribution = Non-parametric tests
14. CharacteristicExposureYesNoYesNoGender (Female)UTIYesNoYes9516No8725Do females ≤18 years experience more UTIs than males?Bivariate Categorical Comparison
15. Pearson’s Chi-Square test: Determines if there is a significant difference between the expected frequencies and the observed frequencies in one or more categories. Fisher’s Exact test: Determines if the proportions of one variable are different depending on the value of the other variable.Small Sample SizesMore exact than Chi-Square test.Bivariate Categorical Comparison
16. Bivariate Comparison of Paired DataPaired t-test: Used for comparison of paired normally distributed continuous data. Example: Comparing pre- vs post-surgical creatinine levels.Wilcoxon signed-rank test: Used for comparison of paired skewed continuous data. Example: Comparing results from pre- vs post-diet weightMcNemar’s test: Used to compare paired categorical data.Example: Comparing yes/no answers on a pre- vs post-training survey.
17. Multivariate Models
18. Analysis of Variance (ANOVA)Compares between three or more groupsTests for association between the groups
19. Independent vs Dependent VariablesIndependent Variable: Not influenced by anything Dependent Variable: Depends on something“(Independent Variable) causes a change in (Dependent Variable) and it isn’t possible that (Dependent Variable) could cause a change in (Independent Variable)”
20. Analysis of Variance (ANOVA)One-way ANOVA: 1 dependent variable and 1 independent variable.Two-way ANOVA: 1 dependent variable and >1 independent variable. AssumptionsA normal distribution.The samples must be independent from each other.The variances in the groups must be equal.
21. Regression AnalysisTests for a relationship between two or more variablesExamines the influence of independent variable on dependent variableCalculates prediction models
22. Linear RegressionSimple Linear Regression: 1 dependent and 1 independentMultiple Linear Regression: 1 dependent and >1 independentAssumptionsVariables compared MUST have a linear relationshipBoth variables MUST be continuous
23. Multiple Linear RegressionAdjusts for confounding effectsCreate prediction models
24. Accuracy of Prediction ModelsTeaching Cohort: Sample where you go the prediction model from.Validation Cohort: Sample population where you test the model.
25. Correlation vs Linear RegressionBoth evaluate for a relationship between two variablesBUTCorrelation doesn’t fit a line through the data pointsCorrelation doesn’t care about cause and effect
26. Logistic Regression AssumptionsDependent variable is binaryData isn’t matched or pairedIndependent variables are not highly correlatedDoesn’t require a linear relationshipDoesn’t require equal variancesOften require larger sample sizes
27. Regression Rule of ThumbOne in Ten Rule (n/10)One predictive variable can be studied for every ten events
28. Analysis that Adjust for Time
29. Kaplan Meier EstimatorNon-parametricEstimated probability of “event free” survival No Hypoplasia HypoplasiaTime PeriodProb SE ProbSE 2.5 years0.9650.0030.6330.0815 years0.9650.0090.5940.0857.5 years0.9650.0090.5940.08510 years0.9650.009590.40.08595% CI for Survival0.633 + 0.081 = 0.7130.633 – 0.081 = 0.552Probability of Death1 - 0.633 = 0.0.3671 - 0.713 = 0.2871 - 0.552 = 0.448
30. Cox Proportional Hazards (Cox Regression)Estimates hazards ratios (risk)Binary dependent variable≥1 independent variables
31. Cox Regression Unadjusted Multivariable ModelInitial Hospitalization CharacteristicHR95% CIp-value HR95% CIp-valueInitial Hospitalization Length of Stay (days)1.21.1 – 1.3<0.0001 1.21.1 – 1.3<0.0001Admit Age (days)1.01.0 - 1.00.061 1.01.0 – 1.00.138<37 week at birth5.52.8 - 11.0<0.0001 2.91.3 – 6.30.008Renal Agenesis7.02.7 – 18.3<0.0001 2.60.9 – 7.40.075Renal Dysplasia2.10.9 – 4.40.055 1.00.5 – 2.40.930Sepsis2.41.0 – 5.90.049 0.80.3 – 2.10.648Pulmonary Hypoplasia13.26.7 – 26.3<0.0001 7.23.1 – 16.8<0.0001Hospital Volume (reference ≤2 PUV case/yr) >2 PUV cases/yr0.350.16 - 0.760.008 0.40.2 – 0.90.034Table III. Risk factors for mortality in unadjusted and multivariable model Cox regression analysis.
32.