Difference between revisions of "Chi-squared test"
| (2 intermediate revisions by one user not shown) | |||
| Line 1: | Line 1: | ||
| − | + | '''Chi-square''' is a [[quantitative research|quantitative number]] calculated by finding the difference between each observed and theoretical frequency for each possible outcome, squaring them, dividing each by the theoretical frequency, and taking the sum of the results. The number of degrees of freedom is equal to the number of possible outcomes, minus 1. | |
| − | + | When working with nominal data like Likert scales that have been grouped into categories, we can arrange them in contingency tables and evaluate them with the chi-square test, first investigated by Karl Pearson. | |
| − | + | ||
| − | + | A chi-square probability of 0.05 or less is commonly interpreted as justification for rejecting the null hypothesis. | |
== Study Design== | == Study Design== | ||
| + | |||
| + | HIDA proposes to analyze all hospital nursing units using a survey study before and after ([[Pretest-Posttest|pretest-post-test design]])the “big-bang” implementation of its new [[EMR|EHR]]. | ||
We propose a cluster sampling of nursing units, with randomized sampling of specific clusters. Face-to-face interviews using a Likert rating scales will focus on the following: | We propose a cluster sampling of nursing units, with randomized sampling of specific clusters. Face-to-face interviews using a Likert rating scales will focus on the following: | ||
| Line 23: | Line 24: | ||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
| − | |||
== Problems== | == Problems== | ||
Latest revision as of 20:38, 30 November 2011
Chi-square is a quantitative number calculated by finding the difference between each observed and theoretical frequency for each possible outcome, squaring them, dividing each by the theoretical frequency, and taking the sum of the results. The number of degrees of freedom is equal to the number of possible outcomes, minus 1.
When working with nominal data like Likert scales that have been grouped into categories, we can arrange them in contingency tables and evaluate them with the chi-square test, first investigated by Karl Pearson.
A chi-square probability of 0.05 or less is commonly interpreted as justification for rejecting the null hypothesis.
Study Design
HIDA proposes to analyze all hospital nursing units using a survey study before and after (pretest-post-test design)the “big-bang” implementation of its new EHR.
We propose a cluster sampling of nursing units, with randomized sampling of specific clusters. Face-to-face interviews using a Likert rating scales will focus on the following:
- ease of data gathering
- ease of data entry/charting
- ease of time management
- disruption of patient care
- efficiency of communication with physicians
- efficiency of communication with pharmacy
- time to pharmacy order arrivals
Data Analysis
Data will be analyzed using Chi Square test of homogeneity to test the null hypothesis that there will be no difference in responses to the survey before and after implementation of the new EHR.
Problems
If the sample size is less than 5, a Yates correction is required, or a Fisher’s exact test is required, which can be arduous, but are usually included in many software packages.
Examples in Medical Informatics:
Most outcome studies will involve chi-square testing if Likert scales are used.
