Azar, normalidad y causalidad Andrés Gempeler
Medición • Parte de los procesos centrales en la construcción del conocimiento cien9fico. • Variables: aquellas caracterísAcas de la realidad que pueden modificarse en el Aempo o según cambios en el espacio.
Variables
“Duras” y “Blandas” • 3 mundos de Karl Popper ó los 3 planos de Roger Penrose – 1: mundo Lsico; material – 2: mundo psicológico, autoconsciencia; senAmientos, pensamientos, memoria. – 3: ideas y cultura; razón y sus instrumentos
Variables
“Duras” y “Blandas” • 1: mundo Lsico; material. • 2: mundo psicológico, autoconsciencia; senAmientos, pensamientos, memoria. • 3: ideas y cultura; razón y sus instrumentos.
– Variables de planos 1 y 3 = “objeAvas” = “duras” Replicables por diferentes observadores.
– Variables de plano 2 = “subjeAvas” = blandas RelaAvas al sujeto que las experimenta.
Variable Una caracterísAca de un ente, de un conjunto de entes, de un fenómeno, o de un conjunto de fenómenos eventos, que puede adoptar diferentes valores (cuanAtaAvas) o estados (cualitaAvas).
Medición La percepción sistemáAca y ordenada de caracterísAcas de la realidad
Medición y fuentes de variabilidad (error) Hay 4 posibles explicaciones para los respuesta que damos a la pregunta de invesAgación luego de hacer mediciones:
1. Es correcta 2. Es incorrecta por errores provenientes de variación aleatoria 3. Es incorrecta por la introducción de errores sistemáAcos 4. Es incorrecta por una combinación de las dos anteriores
Medición y fuentes de variabilidad (error) Mediciones pueden ser correctas o incorrectas. La confianza que tenemos sobre la medición depende de los atributos del método de medición, que son:
– La exacAtud – La precisión
ExacAtud “Accuracy” – Validez En qué grado el resultado de la medición refleja el valor verdadero de una variable Mide la realidad? Validez, y exacAtud, usualmente se refieren a la ausencia de sesgo
Precisión “Precision” – Confiabilidad o Consistencia Rango de variación o fluctuación de mediciones de una misma variable, con un mismo instrumento. Mientras menor sea, hay mayor precisión. Validez, y exacAtud, usualmente se refieren a la ausencia de sesgo
El resultado de la medición Será veraz o correcto A la variabilidad “extraña” que no es inherente al verdadero estado de la variable medida, se le llama Error
Desviación sistemáAca de la realidad
SESGO
Error aleatorio En ausencia de sesgo, aún existe la posibilidad de error, por mala suerte, por azar.
Error aleatorio Muestra a9pica?
Azar -‐ Variabilidad Teoría de probabilidades nos permite esAmar el posible papel del azar en el resultado de una medición específica. (Valor de p)
Azar -‐ Variabilidad Por la incerAdumbre aleatoria inherente a la medición, cuando se obAene un valor, si no está sesgado, se espera que sea compaAble con un intervalo de valores cuyo rango depende de la consistencia (reproducibilidad) del proceso. (intervalo de confianza)
Causalidad
¿Definición? • • • •
Producción Causa necesaria Causas suficiente – componente Causalidad probabilísAca
Postulados de Koch • The microorganism must be found in abundance in all organisms suffering from the disease, but should not be found in healthy organisms. • The microorganism must be isolated from a diseased organism and grown in pure culture. • The cultured microorganism should cause disease when introduced into a healthy organism. • The microorganism must be reisolated from the inoculated, diseased experimental host and idenAfied as being idenAcal to the original specific causaAve agent.
Criterios de Bradford Hill
Medición de variables Análisis à Asociación
¡Asociación no es lo mismo que causalidad!
Otra vez…
¡Asociación no es lo mismo que causalidad!
¡Asociación no es lo mismo que causalidad!
Normalidad / Anormalidad
Normalidad estadísAca
(or “Gaussian,” after the mathematit described it). The normal distribution, tistical theory, describes the frequency
curve is shown in Figure 3.7. The curve is symmetr cal and bell shaped. It has the mathematical proper that about two-thirds of the observations fall with
Frequency
Normalidad estadísAca
Standard deviations –3
–2 2.14
Percent of area under the curve
–1 13.59
0 34.13
+1 34.13
+2 13.59
68.26 95.44 99.72
Figure 3.7 ■ The normal (Gaussian) distribution.
+3 2.14
Normalidad = salud? • Normal como usual, común o esperado. – Peso de los habitantes en PAI – Peso de los habitantes en PBI.
Probablemente ninguno es salud…
heavy house telephone, a period (Fig. declining fu across the B 21% to 38% between no on the health
16
Mortality
8 4 2 1 0 115 120
140
160
180
Usual systolic blood pressure (mm Hg) Figure 3.9 ■ Ischemic heart disease mortality for people ages 40 to 49 years is related to systolic blood pressure throughout the range of values occurring in most people. There is no threshold between normal and abnormal. “Mortality” is presented as a multiple of the baseline rate. (Data from Prospective Studies Collaboration. Age-specific relevance of usual blood pressure to vascular mortality: a meta-analysis of individual data for one million adults in 61 prospective studies. Lancet 2002;360:1903–1913.)
Abnormal Condition Clinical Ou
It makes intuiti or finding as “a a better outcom good sense for dition is causin no difference, w symptomatic p
Mortality rate / 100,000 person-years
A
20,000
16,000
12,000
8,000
4,000 <18.5
18.5–21.9 22.0–24.9 25.0–27.4 27.5–29.9
30.0–34.9
≥35.0
30.0–34.9
≥35.0
2
Body mass index (kg/m )
B
Men with functional decline (%)
40
30
20
10
0 <18.5
18.5–21.9 22.0–24.9 25.0–27.4 27.5–29.9 2
Body mass index (kg/m ) Figure 3.10 ■ Abnormal as associated with disease and other patient outcomes. The relationship between body mass index and (A) total mortality and (B) functional decline in men age 65 and older on Medicare. Body mass index is weight in kilograms divided by height in meters
cance of a ough expete patients’ gainst such estionnaires nicians and y. To overe, researchr states. To he numbers day perforfsky Perfornal capacity ents receivrmal) to 0 e of 60? At sional assiseir personal
All observations are subject to variation because of the performance of the instruments and observers involved in making the measurements. The conditions
Fuentes de variación
Table 3.3 Sources of Variation Source of Variation
Definition
Measurement Variation Instrument
The means of making the measurement
Observer
The person making the measurement
Biologic Variation Within individuals
Changes in a person at different times and situations
Between individuals
Biologic differences from person to person
are used as part Injury and Acute (ALI-ARDS), seterial hypoxemia But do specialists read radiographs rts in pulmonary x-rays from critidecided whether c criteria for the centage of radiodiagnosis ranged experts (Fig. 3.2), ference between
100
Radiographs read positive (%)
ased result (lack of ack of reliability). It variation by making d by following staneasurements involve hines, variation can o control.
80
60
40
20
0
Readings by 21 experts Figure 3.2 ■ Observer variability. Variability among 21 specialists reading chest x-rays for acute lung injury and acute respiratory distress syndrome. The percentage of radiographs read as positive for the diagnosis varied from 36% to 71% among the experts. (Data from Rubenfeld GD, Caldwell E, Granton J, et al. Interobserver variability in applying a radiographic definition for ARDS. Chest 1999;116: 1347–1353.)
CONDITIONS OF MEASUREMENT
DISTRIBUTION OF MEASUREMENT
SOURCE OF VARIATION
Within individual patient Simultaneous–same observer
Measurement
Simultaneous–2 observers
Measurement
Between visits
Biologic
Among patients
Biologic
60
70
80
90
Diastolic blood pressure (mm Hg)
100
110
30 20
Serum potassium
Alkaline phosphatase
20
10
Percent
10
3.0
4.0
5.0
20 40
60
mEq/L
80 100 120 140
Units
30
40
Plasma glucose
20
30
Hemoglobin 20
10 10
100
150
mg/100 mL
200
8
9
10
11
12
13
14
15
16
g/100 mL
Figure 3.6 ■ Actual clinical distributions. (Data from Martin HF, Gudzinowicz BJ, Fanger H. Normal Values in Clinical Chemistry. New York: Marcel Dekker; 1975.)
The Normal Distribution
distribution of repeated measurements of the same
Mensaje de la clase
¡Asociación no es lo mismo que causalidad!
Hasta luego!
Four diVerent types of causal relaAons can be derived from these two definiAons: necessary and suYcient, necessary but not suYcient, suYcient but not necessary, and neither necessary nor suYcient. A small minority of epidemiologists maintain that the term “cause” should be limited to highly specific necessary condiAons.21 22 The view that all causes must be necessary for their eVects is tradiAonally associated with the germ theory of disease,
• Cigareqe • smoke, for example, is not necessary for • development of lung cancer. In fact, some epidemiology • texts and commentaries have stated • that causes of complex chronic diseases, like • cancer and heart disease, tend to fit into the • “neither necessary nor suYcient” category
s. A suYcientcomponent cause is made up of a number of components, no one of which is suYcient for the disease on its own. When all the components are present, however, a suYcient cause is formed. Because more than one set of components • may be suYcient for the same eVect, a • disease may have mulAple causes. • • • • • •
hold in the absence of empirical evidence. In short, the suYcientcomponent cause definiAon requires that we assume the existence of countless hidden eVect • modifiers to turn every less than perfect correlaAon • into pure determinism • • • •