7.3
Intervalles de confiance
Une estimation ponctuelle sans biais est souvent insuffisante pour prédire un paramètre de la population, tel que la moyenne ou la proportion d'une po…
Une estimation ponctuelle non biaisée n’est pas toujours suffisante pour prédire avec précision un paramètre de population donné, par exemple la proportion ou la moyenne de la population.
Ainsi, pour obtenir un meilleur jugement d’un paramètre de population, une plage de valeurs peut être tirée de la distribution des données d’échantillon pour estimer la valeur réelle du paramètre de population.
Cette plage est appelée estimation d’intervalle, plus communément appelée intervalle de confiance.
Contrairement à l’estimation ponctuelle, l’intervalle de confiance génère une plage de valeurs à l’intérieur de deux limites, l’une inférieure et l’autre supérieure, généralement appelées limites de confiance.
L’intervalle de confiance de la proportion de la population peut être représenté en écrivant la limite inférieure calculée, suivie de la proportion de la population, puis de la limite supérieure calculée.
Dans cette équation, 
est la proportion de la population, et E est la marge d’erreur.
En termes plus simples, il peut également être exprimé comme suit :
View the full transcript and gain access to JoVE Core videos
Q1: Why is a point estimate alone insufficient for predicting a population parameter?
A point estimate provides only a single value and cannot account for sampling variability or uncertainty. A confidence interval addresses this limitation by generating a range of values with lower and upper confidence limits, offering a more accurate judgment of the true population parameter and indicating the uncertainty in the estimate.
Q2: What are confidence limits and how do they relate to a confidence interval?
Confidence limits are the lower and upper boundaries of a confidence interval, calculated by adding and subtracting the margin of error from a point estimate. These limits define the range within which the true population parameter is expected to fall, with the interval represented as lower limit, point estimate, and upper limit.
Q3: How does the width of a confidence interval affect the accuracy of an estimate?
A narrower confidence interval indicates greater accuracy in predicting the true population parameter. Conversely, a wider interval reflects greater uncertainty. The relationship is inverse: as the confidence interval narrows, the precision and reliability of the point estimate in predicting the actual population parameter increases significantly.
Q4: What is the margin of error and how is it used to calculate confidence limits?
The margin of error (E) quantifies the uncertainty around a point estimate. Confidence limits are calculated by subtracting the margin of error from the point estimate to obtain the lower limit and adding it to obtain the upper limit. This creates the range: point estimate ± E, defining the confidence interval boundaries.
Q5: How does confidence level relate to confidence intervals?
A confidence level indicates the probability that a confidence interval contains the true population parameter. Common confidence levels are 90%, 95%, and 99%. Higher confidence levels produce wider intervals, reflecting greater certainty that the interval captures the actual population parameter value.
Q6: What is the difference between a confidence interval for population proportion and population mean?
Both use the same interval structure with lower and upper limits, but they estimate different parameters. A confidence interval for population proportion uses sample proportion as the point estimate, while a confidence interval for estimating population mean uses sample mean. Both incorporate margin of error to establish their respective ranges.
Q7: How does sample data distribution contribute to creating a confidence interval?
A confidence interval is derived from the sample data distribution to estimate the true population parameter value. By drawing a range of values from this distribution and applying the margin of error, researchers establish confidence limits that reflect both the sample estimate and the uncertainty inherent in sampling variability.