Algumas propostas para imputação de dados faltantes em Teoria de Resposta ao Item
In this work we have studied two imputation procedures for missing data when fitting the three parameters model in item response theory studies. The first proposed method uses as the probability of imputation of correct response the value obtained from the logistic regression of the correct (1) and...
| Egile nagusia: | |
|---|---|
| Beste egile batzuk: | |
| Formatua: | Dissertação |
| Hizkuntza: | portugesa |
| Argitaratua: |
Universidade de Brasília
2024
|
| Gaiak: | |
| Sarrera elektronikoa: | https://hdl.handle.net/20.500.14135/931 |
| Gaia: | In this work we have studied two imputation procedures for missing data when fitting the three parameters model in item response theory studies. The first proposed
method uses as the probability of imputation of correct response the value obtained from the logistic regression of the correct (1) and incorrect (0) answers as a function
of the difficulty parameter estimates for the items with responses. We take the fitted logisitic curve and calculate the probability of imputation of positive response as
a function of the value of the difficulty parameter estimates for the non responded items. The second method is similar to the first one, but we use the probability of
imputation of a positive response provided by the kernel smoothed isotonic regression obtained taking the probability of positive response as a decreasing function of the
estimated difficulty parameter for each item with a response. In a simulation study, the estimates of the item parameters and proficiency were compared to the real values
and also to the estimates obtained for the data with no missing responses. |
|---|