|
Location:
|
HC 3023 & 3023A |
|
Date:
|
Monday, Aug.1 |
|
Time:
|
1:50 PM -2:00 PM
|
|
Author:
|
Li Chen
School of Electronic science and engeering, Southeast University
614-292-2450, chenli.seu@163.com
|
|
Co-Author(s):
|
Jing Han , Liangyu Peng , Yan Tu , Lei Bao
|
|
Abstract:
|
Item Response Theory is a useful tool for analyzing quantitative data. The sample size will impact the uncertainty of the estimated parameters. It is then important to find out the approximate minimum sample size, with which reliable results can be calculated. In this study, we choose R (with its LTM package) to estimate the parameters with different sample sizes, which are randomly selected from the college students' FCI data collected at The Ohio State University. The total number of the data is 3139. The results show an exponential relationship between sample size and the mean difference of the results obtained with subsets of the data. When sample size is larger than 1600, the difference is tolerable for most items and the mean total difference can be controlled within 5%. This can provide useful guide for future data analysis using IRT.
|
|
Footnotes:
|
Supported in part by NIH Award RC1RR028402 and NSF Awards DUE-0633473 and DUE-1044724
|
|
|
|