|
How To... > Work with Statistical Objects > Estimate Parameters and Compute Confidence Intervals
Imagine you’re working with a random sample from some population, such as a sample of voters, a sample of patients with high blood pressure, or a sample of elementary schools in a state. You measure something about the sample, for example, the proportion planning to vote for a certain candidate, the mean blood pressure, or the difference in proportion of students passing a test in two groups. The thing you measure is an estimate of what is true for the population, in other words, a population . But the single number is not very useful. Much more useful is the range of values in which you think the population parameter lies and how confident you are about that range.
Fathom’s objects are designed to compute both the estimate and its confidence interval.
Attribute types
|
Estimate types
|
One numeric attribute (e.g., height)
|
Estimate the Population Mean from Raw Data. Confidence interval is from the one-sample t procedure.
|
Two numeric attributes (e.g., age and income)
|
Estimate Difference of Means from Raw Data. Estimates the difference of means between the two attributes and gives the confidence interval for this difference.
|
One categorical attribute (e.g., sex: "M" or "F")
|
Estimate the Population Proportion from Raw Data. Confidence interval is exact binomial up to a point, then a normal approximation.
|
One categorical attribute with two categories (e.g., sex: "M" or "F") and one numeric attribute (e.g., height)
|
Estimate Difference of Means from Raw Data. Estimates the difference of means for the numeric attribute between the two groups given by the categorical attribute and gives the confidence interval for this difference.
|
Two categorical attributes, one describing a group (e.g. sex: "M" or "F") and the other providing a characteristic of that group (e.g., Race: "white" or "nonwhite")
Two categorical attributes, each one describing a group (e.g., PovertyY1: "Y" or "N" and PovertyY2: "Y" or "N")
|
Estimate Difference of Proportions from Raw Data. Estimates confidence interval for the difference of proportions having a specified category for each attribute. Uses a normal approximation.
|
|
