心理科学
心理科學
심이과학
Psychological Science
2013年
3期
716~721
,共null页
因果推理 取样大小 Suppoa模型 SS效力模型
因果推理 取樣大小 Suppoa模型 SS效力模型
인과추리 취양대소 Suppoa모형 SS효력모형
causal inference, sample size, Support -model, SS -model
使用纸笔测验探讨表格集中呈现信息条件下取样大小对单一因果关系强度推理的影响,并比较五种模型:△P、效力PC、SS效力、Support和χ2的预测与实验数据的相关。结果显示:(1)取样大小对不同的问题有不同的影响,高取样在△P=0时导致了低估汁值,在0〈|△P|=PPC时没有效果,在|△P|〈PPC时有不一致的效果,4倍取样比2倍取样有更明显的取样大小效应;(2)取样大小效应更一致地出现在PPC与|△P|不相等且PPC值较高的问题中;(3)虽然Suppoa模型和ss效力模型都能预测取样大小因素在所有问题上的作用方式,但取样大小效应并不如因果支持模型所预测的那么强大,在两种因果方向条件下SS效力模型都能最好地预测被试的反应。
使用紙筆測驗探討錶格集中呈現信息條件下取樣大小對單一因果關繫彊度推理的影響,併比較五種模型:△P、效力PC、SS效力、Support和χ2的預測與實驗數據的相關。結果顯示:(1)取樣大小對不同的問題有不同的影響,高取樣在△P=0時導緻瞭低估汁值,在0〈|△P|=PPC時沒有效果,在|△P|〈PPC時有不一緻的效果,4倍取樣比2倍取樣有更明顯的取樣大小效應;(2)取樣大小效應更一緻地齣現在PPC與|△P|不相等且PPC值較高的問題中;(3)雖然Suppoa模型和ss效力模型都能預測取樣大小因素在所有問題上的作用方式,但取樣大小效應併不如因果支持模型所預測的那麽彊大,在兩種因果方嚮條件下SS效力模型都能最好地預測被試的反應。
사용지필측험탐토표격집중정현신식조건하취양대소대단일인과관계강도추리적영향,병비교오충모형:△P、효력PC、SS효력、Support화χ2적예측여실험수거적상관。결과현시:(1)취양대소대불동적문제유불동적영향,고취양재△P=0시도치료저고즙치,재0〈|△P|=PPC시몰유효과,재|△P|〈PPC시유불일치적효과,4배취양비2배취양유경명현적취양대소효응;(2)취양대소효응경일치지출현재PPC여|△P|불상등차PPC치교고적문제중;(3)수연Suppoa모형화ss효력모형도능예측취양대소인소재소유문제상적작용방식,단취양대소효응병불여인과지지모형소예측적나요강대,재량충인과방향조건하SS효력모형도능최호지예측피시적반응。
Participant's judgment on causal strength can vary with sample size of contingency. Griffiths & Tenenbaum (2005) show that sample size can affect participant's judgment on contingency with △P = 0. But Lu et al. (2008) found that sample size has an effect on contingency with △P 〉 0 in the opposite direction, and high power - PC can offset low sample size. Then, what is the effect of sample size on contingencies with different △P and Power- PC? The research conducted an experiment to investigate whether participant's cansal judgment of seven contingencies vary with sample size. Contingency was presented on a booklet with a table format, and participants gave their causal - strength judgments under each contingency. Results show that sample size factor has different effects on different contingencies : high sample size leads to low estimate on contingency with AP = 0, but has no effect on contingency with 0 〈 |△P | = power - PC, but leads to a high estimate on contingency with | △P | 〈 power - PC = 1. Four times sample has more effect than two times sample. It seems that sample size effect is apt to emerge from the contingencies with | △P | 〈 power - PC. Meanwhile, a comparison of predictions of five models, △P, power - PC, SS, Support and χ2, was conducted. Results indicated that support and )(2 model gave some predictions contradicting participants'judgment, including lower predict on contingency with | △P | = . 33 and power - PC = . 50 than contingency with | △P | = . 33 and power - PC --- . 33, and lower predict on contingency with | △P | . 33 and power - PC = 1 than contingency with | △P | = power - PC = . 67. Support and SS model could predict sample size effect on each contingency, but the effectiveness of sample size was not so powerful as the prediction of the support model, while SS was the best model to predict the participants" performance in two kinds of causal direction.