Introduction to Logistic Regression Rachid Salmi, Jean-Claude Desenclos, Thomas Grein, Alain Moren

Oral contraceptives (OC) and myocardial infarction (MI) Case-control study, unstratified data OC MIControlsOR Yes No Ref. Total

Oral contraceptives (OC) and myocardial infarction (MI) Case-control study, unstratified data Smoking MIControlsOR Yes No Ref. Total

Odds ratio for OC adjusted for smoking = 4.5

Number of cases One case Days Cases of gastroenteritis among residents of a nursing home, by date of onset, Pennsylvania, October 1986

ProteinTotalCasesAR%RR suppl. YES NO Total Cases of gastroenteritis among residents of a nursing home according to protein supplement consumption, Pa, 1986

Sex-specific attack rates of gastroenteritis among residents of a nursing home, Pa, 1986 SexTotalCases AR(%)RR & 95% CI Male Reference Female ( ) Total

Attack rates of gastroenteritis among residents of a nursing home, by place of meal, Pa, 1986 MealTotal CasesAR(%)RR & 95% CI Dining room Reference Bedroom ( ) Total

Age – specific attack rates of gastroenteritis among residents of a nursing home, Pa, 1986 Age groupTotalCasesAR(%) Total

Attack rates of gastroenteritis among residents of a nursing home, by floor of residence, Pa, 1986 FloorTotalCasesAR (%) One Two Three Four Total

Multivariate analysis Multiple models –Linear regression –Logistic regression –Cox model –Poisson regression –Loglinear model –Discriminant analysis – Choice of the tool according to the objectives, the study, and the variables

Simple linear regression Table 1 Age and systolic blood pressure (SBP) among 33 adult women

SBP (mm Hg) Age (years) adapted from Colton T. Statistics in Medicine. Boston: Little Brown, 1974

Simple linear regression Relation between 2 continuous variables (SBP and age) Regression coefficient 1 –Measures association between y and x –Amount by which y changes on average when x changes by one unit –Least squares method y x Slope

Multiple linear regression Relation between a continuous variable and a set of i continuous variables Partial regression coefficients i –Amount by which y changes on average when x i changes by one unit and all the other x i s remain constant –Measures association between x i and y adjusted for all other x i Example –SBP versus age, weight, height, etc

Multiple linear regression Predicted Predictor variables Response variable Explanatory variables Outcome variable Covariables Dependent Independent variables

Logistic regression (1) Table 2 Age and signs of coronary heart disease (CD)

How can we analyse these data? Compare mean age of diseased and non-diseased –Non-diseased: 38.6 years –Diseased: 58.7 years (p<0.0001) Linear regression?

Dot-plot: Data from Table 2

Logistic regression (2) Table 3 Prevalence (%) of signs of CD according to age group

Dot-plot: Data from Table 3 Diseased % Age group

Logistic function (1) Probability of disease x

Transformation logit of P(y|x) { = log odds of disease in unexposed = log odds ratio associated with being exposed e = odds ratio

Fitting equation to the data Linear regression: Least squares Logistic regression: Maximum likelihood Likelihood function –Estimates parameters and –Practically easier to work with log-likelihood

Maximum likelihood Iterative computing –Choice of an arbitrary value for the coefficients (usually 0) –Computing of log-likelihood –Variation of coefficients values –Reiteration until maximisation (plateau) Results –Maximum Likelihood Estimates (MLE) for and –Estimates of P(y) for a given value of x

Multiple logistic regression More than one independent variable –Dichotomous, ordinal, nominal, continuous … Interpretation of i –Increase in log-odds for a one unit increase in x i with all the other x i s constant –Measures association between x i and log-odds adjusted for all other x i

Statistical testing Question –Does model including given independent variable provide more information about dependent variable than model without this variable? Three tests –Likelihood ratio statistic (LRS) –Wald test –Score test

Likelihood ratio statistic Compares two nested models Log(odds) = + 1 x x x 3 (model 1) Log(odds) = + 1 x x 2 (model 2) LR statistic -2 log (likelihood model 2 / likelihood model 1) = -2 log (likelihood model 2) minus -2log (likelihood model 1) LR statistic is a 2 with DF = number of extra parameters in model

Coding of variables (2) Nominal variables or ordinal with unequal classes: –Tobacco smoked: no=0, grey=1, brown=2, blond=3 –Model assumes that OR for blond tobacco = OR for grey tobacco 3 –Use indicator variables (dummy variables)

Indicator variables: Type of tobacco Neutralises artificial hierarchy between classes in the variable "type of tobacco" No assumptions made 3 variables (3 df) in model using same reference OR for each type of tobacco adjusted for the others in reference to non-smoking

Reference Hosmer DW, Lemeshow S. Applied logistic regression. Wiley & Sons, New York, 1989

Logistic regression Synthesis

Salmonella enteritidis Protein supplement S. Enteritidis gastroenteritis Sex Floor Age Place of meal Blended diet

Unconditional Logistic Regression Term Odds Ratio 95% C.I.Coef.S. E. Z- Statistic P- Value AGG (2/1)1,67950,263410,70820,51850,94520,54860,5833 AGG (3/1)1,75700,32499,50220,56360,86120,65450,5128 Blended (Yes/No)1,03450,32773,26600,03390,58660,05780,9539 Floor (2/1)1,61260,26759,72200,47780,91660,52130,6022 Floor (3/1)0,72910,09915,3668-0,31591,0185-0,31020,7564 Floor (4/1)1,11370,15737,88700,10760,99880,10780,9142 Meal1,59420,49535,13170,46640,59650,78190,4343 Protein (Yes/No)9,09183,021927,35332,20740,56203,92780,0001 Sex1,30240,22787,44680,26420,88960,29700,7665 CONSTANT***-3,00802,0559-1,46310,1434

Unconditional Logistic Regression TermOdds Ratio 95% C.I.CoefficientS. E.Z-StatisticP-Value Age1,02340,96601,08420,02310,02940,78480,4326 Blended (Yes/No)1,01840,32203,22070,01830,58740,03110,9752 Floor (2/1)1,64400,27459,84680,49710,91330,54430,5862 Floor (3/1)0,71320,09725,2321-0,33791,0167-0,33240,7396 Floor (4/1)1,07080,15227,53220,06840,99530,06870,9452 Meal1,65610,52365,23790,50450,58750,85870,3905 Protein (Yes/No)8,76782,952126,04032,17110,55543,90910,0001 Sex1,19570,21356,69810,17870,87910,20330,8389 CONSTANT***-4,28962,8908-1,48390,1378

Logistic Regression Model Summary Statistics ValueDFp-value Deviance107, Likelihood ratio test34,80688< Parameter Estimates 95% C.I. TermsCoefficientStd.Errorp-valueORLowerUpper %GM-1,88571,04200,07030,15170,01971,1695 SEX ='2'0,21390,88120,80821,23850,22026,9662 FLOOR ='2'0,49870,90830,58291,64660,27769,7659 ²FLOOR ='3'-0,32351,01500,75000,72360,09905,2909 FLOOR ='4'0,10880,98390,91191,11500,16217,6698 MEAL ='2'0,53080,56130,34431,70020,56595,1081 Protein ='1'2,18090,5303< ,85413,131625,034 TWOAGG ='2'0,19040,51620,71221,20980,43993,3272 Termwise Wald Test TermWald Stat.DFp-value FLOOR1,081230,7816

Poisson Regression Model Summary Statistics ValueDFp-value Deviance60, Likelihood ratio test67,73788< Parameter Estimates 95% C.I. TermsCoefficientStd.Errorp-valueRRLowerUpper %GM-1,82130,84460,03100,16180,03090,8471 SEX ='2'0,12950,71060,85541,13830,28274,5828 FLOOR ='2'0,25030,68670,71541,28440,33444,9343 FLOOR ='3'-0,14220,80320,85950,86740,17974,1877 FLOOR ='4'0,13680,72630,85061,14660,27614,7608 MEAL ='2'0,23730,38540,53811,26780,59562,6987 Protein ='1'1,06580,34130,00182,90321,48715,6679 TWOAGG ='2'0,06450,36820,86111,06660,51822,1951 Termwise Wald Test TermWald Stat.DFp-value FLOOR0,417830,9365

Cox Proportional Hazards TermHazard Ratio95%C.I.CoefficientS. E.Z-StatisticP-Value _AGG (2/1)1,06660,51832,1950,06450,36820,1750,8611 Floor(2/1)1,28440,33444,93420,25030,68670,36460,7154 Floor(3/1)0,86740,17974,1876-0,14220,8032-0,1770,8595 Floor(4/1)1,14660,27614,76070,13680,72630,18830,8506 Meal (2/1)1,26780,59572,69860,23730,38540,61570,5381 Protein(Yes/No)2,90321,48715,66781,06580,34133,12250,0018 Sex (2/1)1,13830,28274,58270,12950,71060,18220,8554 Convergence:Converged Iterations:5 -2 * Log-Likelihood:346,0200 TestStatisticD.F.P-Value Score17,172770,0163 Likelihood Ratio15,488970,0302