不同参照基准与年级离散程度对垂直等值的影响研究
发布时间：2019-05-24 11:44
【摘要】：本文通过蒙特卡洛模拟方法，采用锚题设计，模拟不同参照基准与年级离散程度下，对垂直等值产生的影响。结果发现： （1）在均值上，以年级1为基准的各年级的难度均值与1年级距离越远，其值与真值的差异越大；同时随着年级间的离散程度增大，所有年级的难度的D值绝对值呈现明显的递增趋势。在基准年级2上，只有年级4呈现递增趋势，但是相比较基准年级1，在年级2、年级3与年级4上D值的绝对值明显下降。在锚题上，基准年级为1年级时，在1与1.5的年级离散上，锚题难度值随着年级增大，真值与估计值的差异呈现递增趋势；同时不同年级间锚题随着年级离散程度增大，与真值差异有规律的增大；在2年级的基准上则只有Mm34b随着效应呈现递增趋势，，同时与1年级基准相比，其与真值差异明显要更小。 （2）在b参数RMSE上，难度b值在1年级的基准上，1与2年级随着离散程度增大仍然比较稳定；而3与4年级趋近，但是随着年级离散程度增大而明显增大，同时在不同效应大小上普遍高于年级1与年级2；在2年级基准上，年级1与年级2、年级3趋近，且随着年级离散趋势增大RMSE的变化不明显，但4年级随着离散增大明显上升，且RMSE普遍高于其他年级。 （3）在a参数RMSE上，无论是参照基准年级1还是参照基准年级2，年级1、年级2与年级3上的区分度RMSE在不同效应上趋于稳定与一致，但是年级4则随着年级离散程度增大而增大，且同一离散程度上普遍高于其他年级。 （4）θ参数RMSE值，其与b值的RMSE值基本趋于一致，只是年级3与年级4在不同年级离散程度上起伏程度相对较小。 （5）效应大小ES值，在ES估计与真值的对比上，只有基准年级1的ES值随着年级与年级离散程度增大而增大，但只呈现微弱的递增趋势，而基准年级2只在3与4年级间，随着效应大小增大呈现递增趋势，其余年级间都呈现无规律现象，且基准年级2比基准年级1的D值的绝对值更小. （6）水平距离HD的均值，HD均值在与真值比较上，无论是基准1或者基准2，随着年级与年级之间的离散程度ES增加，HD与真值的差呈现明显的递增趋势，但相对而言，基准年级2的D值的绝对更小。 （7）不同比例水平距离HD值，不同的参照基准上，在0.5效应大小上，不同年级与不同百分比的水平距离上的D值没有明显的规律变化，但是随着效应大小的增大，越远离基准年级，D值呈现规律性的增加，且不同百分比上的水平距离也呈现规律性的增大，但是总体上，基准年级2明显比基准年级1的精度更好。 综合上述认为，不同基准与年级离散程度会对垂直等值效果产生较大的影响。
[Abstract]:In this paper, Monte Carlo simulation method is used to simulate the influence of different reference frames and grade discretization on vertical equivalence by using anchor design. The results show that: (1) in the mean value, the farther the distance between the average difficulty of each grade based on grade 1 and grade 1, the greater the difference between the value and the true value; At the same time, with the increase of the degree of dispersion between grades, the absolute value of D value of difficulty in all grades shows an obvious increasing trend. In grade 2, only grade 4 showed an increasing trend, but compared with grade 1, the absolute value of D value in grade 2, grade 3 and grade 4 decreased significantly. On the anchor problem, when the benchmark grade is grade 1, the difficulty value of anchor problem increases with the grade of grade 1 and 1.5, and the difference between the true value and the estimated value shows an increasing trend. At the same time, with the increase of the degree of dispersion of different grades, the difference between anchoring problems and true values increases regularly. On the benchmark of grade 2, only Mm34b showed an increasing trend with the effect, and compared with the benchmark of grade 1, the difference between the benchmark and the true value was significantly smaller. (2) on the b parameter RMSE, the difficulty b value is on the benchmark of grade 1, and the grade 1 and grade 2 are still stable with the increase of the degree of discretization; Grade 3 and grade 4 approach, but it increases obviously with the increase of grade dispersion, and is generally higher than grade 1 and grade 2 in different effect sizes. On the benchmark of grade 2, grade 1 and grade 2 and grade 3 approached, and the change of RMSE was not obvious with the increase of grade discretization, but the RMSE of grade 4 increased obviously with the increase of dispersion, and the RMSE was generally higher than that of other grades. (3) on the a parameter RMSE, the discrimination RMSE on the reference grade 1 or the reference grade 2, grade 1, grade 2 and grade 3 tends to be stable and consistent in different effects. However, grade 4 increases with the increase of grade discretization, and the same degree of discretization is generally higher than that of other grades. (4) the RMSE value of theta parameter tends to be basically consistent with the RMSE value of b value, but the fluctuation degree of grade 3 and grade 4 is relatively small in different grades. (5) in the comparison of ES estimation and true value, only the es value of benchmark grade 1 increases with the increase of grade and grade dispersion, but only shows a weak increasing trend, while the base grade 2 is only between grade 3 and grade 4. With the increase of the effect, there is an irregular phenomenon among the other grades, and the absolute value of the D value of the benchmark grade 2 is smaller than that of the benchmark grade 1. (6) the mean value of horizontal distance HD and the mean value of HD are compared with the true value, whether it is benchmark 1 or benchmark 2, with the increase of ES between grade and grade, the difference between HD and true value shows an obvious increasing trend, but relatively speaking, The D value of benchmark grade 2 is absolutely smaller. (7) there is no obvious regular change in D value between different grades and different percentages of horizontal distance on different proportion horizontal distance HD values and different reference frames, but with the increase of effect size, the D value of different grades and different percentages of horizontal distance does not change regularly, but with the increase of effect size, the D value of different grades and different percentages of horizontal distance does not change regularly, but with the increase of effect size, The farther away from the base grade, the D value increases regularly, and the horizontal distance in different percentages also increases regularly, but on the whole, the accuracy of benchmark grade 2 is obviously better than that of reference grade 1. According to the above, the dispersion degree of different datum and grade will have a great influence on the vertical equivalence effect.
【学位授予单位】：江西师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：B841
本文编号：2484832
[Abstract]:In this paper, Monte Carlo simulation method is used to simulate the influence of different reference frames and grade discretization on vertical equivalence by using anchor design. The results show that: (1) in the mean value, the farther the distance between the average difficulty of each grade based on grade 1 and grade 1, the greater the difference between the value and the true value; At the same time, with the increase of the degree of dispersion between grades, the absolute value of D value of difficulty in all grades shows an obvious increasing trend. In grade 2, only grade 4 showed an increasing trend, but compared with grade 1, the absolute value of D value in grade 2, grade 3 and grade 4 decreased significantly. On the anchor problem, when the benchmark grade is grade 1, the difficulty value of anchor problem increases with the grade of grade 1 and 1.5, and the difference between the true value and the estimated value shows an increasing trend. At the same time, with the increase of the degree of dispersion of different grades, the difference between anchoring problems and true values increases regularly. On the benchmark of grade 2, only Mm34b showed an increasing trend with the effect, and compared with the benchmark of grade 1, the difference between the benchmark and the true value was significantly smaller. (2) on the b parameter RMSE, the difficulty b value is on the benchmark of grade 1, and the grade 1 and grade 2 are still stable with the increase of the degree of discretization; Grade 3 and grade 4 approach, but it increases obviously with the increase of grade dispersion, and is generally higher than grade 1 and grade 2 in different effect sizes. On the benchmark of grade 2, grade 1 and grade 2 and grade 3 approached, and the change of RMSE was not obvious with the increase of grade discretization, but the RMSE of grade 4 increased obviously with the increase of dispersion, and the RMSE was generally higher than that of other grades. (3) on the a parameter RMSE, the discrimination RMSE on the reference grade 1 or the reference grade 2, grade 1, grade 2 and grade 3 tends to be stable and consistent in different effects. However, grade 4 increases with the increase of grade discretization, and the same degree of discretization is generally higher than that of other grades. (4) the RMSE value of theta parameter tends to be basically consistent with the RMSE value of b value, but the fluctuation degree of grade 3 and grade 4 is relatively small in different grades. (5) in the comparison of ES estimation and true value, only the es value of benchmark grade 1 increases with the increase of grade and grade dispersion, but only shows a weak increasing trend, while the base grade 2 is only between grade 3 and grade 4. With the increase of the effect, there is an irregular phenomenon among the other grades, and the absolute value of the D value of the benchmark grade 2 is smaller than that of the benchmark grade 1. (6) the mean value of horizontal distance HD and the mean value of HD are compared with the true value, whether it is benchmark 1 or benchmark 2, with the increase of ES between grade and grade, the difference between HD and true value shows an obvious increasing trend, but relatively speaking, The D value of benchmark grade 2 is absolutely smaller. (7) there is no obvious regular change in D value between different grades and different percentages of horizontal distance on different proportion horizontal distance HD values and different reference frames, but with the increase of effect size, the D value of different grades and different percentages of horizontal distance does not change regularly, but with the increase of effect size, the D value of different grades and different percentages of horizontal distance does not change regularly, but with the increase of effect size, The farther away from the base grade, the D value increases regularly, and the horizontal distance in different percentages also increases regularly, but on the whole, the accuracy of benchmark grade 2 is obviously better than that of reference grade 1. According to the above, the dispersion degree of different datum and grade will have a great influence on the vertical equivalence effect.
【学位授予单位】：江西师范大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：B841
【参考文献】
相关期刊论文 前3条
1 王烨晖;边玉芳;;构建学业发展性量表——垂直等值的应用[J];中国考试;2010年10期
2 蔡艳;丁树良;涂冬波;;铆题比例对等值精度的影响[J];心理学探新;2009年02期
3 王烨晖;边玉芳;辛涛;;垂直等值的应用及最新发展述评[J];心理学探新;2011年05期
本文编号：2484832
本文链接：https://www.wllwen.com/shekelunwen/xinlixingwei/2484832.html
