What Are Confidence Intervals And P-values?

Article

Jonathan Rodriguez Castillo

What are confidence intervals and p-values?

Introducing

The answer to the question posed by the title is clear: it is not only possible to know the statistical significance of a result, but also allow us to evaluate the clinical significance, its practical importance. Possibly for this reason, for many years, the use of confidence intervals (CI) has been strongly recommended by the directors and editors of major medical journals as a way to supplement the results of a study, either accompanying the p value or even as its sustituto2, 3. Currently, it is possible that many journals of high impact factor did not permit post a fundamental result only matched by its value p.

________________________________________

C

onfidence intervals

Inference process is that by which we want to estimate the value of a parameter from the value of a statistician. This estimate can be punctual or by interval. The best estimate of a parameter is simply the corresponding statistical value, but is uninformative because the probability of not finding the correct value is very high, that is why it is customary to give an interval estimate, which is expect to find the parameter value with high probability. This estimate is called the estimation using confidence intervals.

The confidence interval estimation is to determine a possible value range or interval (a, b), which, with a certain probability, contain limits the parameter value population we're looking for. For each sample to obtain a different range for the X% of them will contain the true value of the parameter. This interval is called a confidence interval. Obviously this technique need not always give a correct result, as we have said for some samples the corresponding interval will contain the true value of the parameter and not others. In the likelihood that we right in saying that the interval contains the parameter is called confidence level (or just trust). Also called significance level of the

Likelihood of error in this statement, the significance (probability of error in our interval) is equal to 1 - (confidence level), as the confidence level corresponds to the probability that the interval contains the true parameter value.

C

ontrast of hypothesis

hypothesis tests are able to respond to specific questions we can make on the population parameters of interest, for example: The average daily salt intake by hypertensive is greater than that ingested by people with normal blood pressure?, Does the body temperature of patients who have suffered certain bacterial infection is higher than 36.7 degrees Celsius?, does the proportion of diabetics with vision problems is higher than in the general population?. It is clear that a mechanism able to answer questions like the above would

...