SEM(1)
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">最终</span>到了结构方程模型的分享,听这个名字<span style="color: black;">好似</span>很难懂,<span style="color: black;">不外</span>在<span style="color: black;">实质</span>的理解上<span style="color: black;">倘若</span>你仔细阅读<span style="color: black;">每一个</span>SEM的分享,<span style="color: black;">必定</span>会豁然开朗,<span style="color: black;">这次</span>分享为了让<span style="color: black;">大众</span><span style="color: black;">认识</span>SEM的前世今生,让<span style="color: black;">大众</span>彻底学会用SEM,<span style="color: black;">咱们</span>大<span style="color: black;">大概</span>十次<span style="color: black;">上下</span>的分享,循序渐进。<span style="color: black;">这次</span>分享的内容<span style="color: black;">都数</span>为</span><span style="color: black;">罗胜强教授的《管理学问卷调查<span style="color: black;">办法</span>》</span><span style="color: black;">这本书中<span style="color: black;">针对</span>结构方程模型的解释,这本书可谓是万卷书法系列中最清晰,最容易上手且干货慢慢的一本书,尤其<span style="color: black;">举荐</span>管理学大类的<span style="color: black;">朋友</span>进行学习。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">首要</span>,<span style="color: black;">咱们</span>搜索一下结构方程模型,找到一个<span style="color: black;">关联</span>的<span style="color: black;">诠释</span>:</span>结构方程模型(Structural Equation Modeling,SEM) 是社会科学<span style="color: black;">科研</span>中的一个非常好的<span style="color: black;">办法</span>。该<span style="color: black;">办法</span>在20世纪80年代就<span style="color: black;">已然</span>成熟,<span style="color: black;">可惜国内<span style="color: black;">认识</span>的人并不多</span>。"在社会科学以及经济、市场、管理等<span style="color: black;">科研</span><span style="color: black;">行业</span>,有时需处理多个<span style="color: black;">原由</span>、多个结果的关系,<span style="color: black;">或</span>会碰到不可直接观测的变量<span style="color: black;">(即潜变量)</span>,这些都是传统的统计<span style="color: black;">办法</span><span style="color: black;">不可</span>很好<span style="color: black;">处理</span>的问题。20世纪80年代<span style="color: black;">败兴</span>,结构方程模型<span style="color: black;">快速</span>发展,弥补了传统统计<span style="color: black;">办法</span>的不足,<span style="color: black;">作为</span>多元数据分析的重要工具。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">搜索出的结果中有以下五点优点</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(一)<span style="color: black;">同期</span>处理</span><span style="color: black;">多个</span><span style="color: black;">因变量</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">结构方程分析可<span style="color: black;">同期</span><span style="color: black;">思虑</span>并处理多个因变量。在回归分析或路径分析中,就算统计结果的图表中展示多个因变量,其实在计算回归系数或路径系数时,仍是对<span style="color: black;">每一个</span>因变量逐一计算。<span style="color: black;">因此</span>图表看似对多个因变量<span style="color: black;">同期</span><span style="color: black;">思虑</span>,但在计算对某一个因变量的影响或关系时,都忽略了其他因变量的存在及其影响。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(二)容许自变量和因变量含<span style="color: black;">测绘</span>误差</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">态度、<span style="color: black;">行径</span>等变量,<span style="color: black;">常常</span>含有误差,<span style="color: black;">亦</span><span style="color: black;">不可</span>简单地用单一指标<span style="color: black;">测绘</span>。结构方程分析容许自变量和因变量均含<span style="color: black;">测绘</span>误差。变量<span style="color: black;">亦</span>可用多个指标<span style="color: black;">测绘</span>。用传统<span style="color: black;">办法</span>计算的<span style="color: black;">潜变量</span>间<span style="color: black;">关联</span>系数,与用结构方程分析计算的潜变量间<span style="color: black;">关联</span>系数,可能相差很大。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(三)<span style="color: black;">同期</span>估计因子结构和因子关系</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">假设要<span style="color: black;">认识</span>潜变量之间的<span style="color: black;">关联</span>,<span style="color: black;">每一个</span>潜变量者用多个指标或题目<span style="color: black;">测绘</span>,一个常用的做法是对<span style="color: black;">每一个</span>潜变量先用因子分析计算潜变量(即因子)与题目的关系(即因子负荷),<span style="color: black;">从而</span>得到因子得分,<span style="color: black;">做为</span>潜变量的观测值,<span style="color: black;">而后</span>再计算因子得分,<span style="color: black;">做为</span>潜变量之间的<span style="color: black;">关联</span>系数。这是两个独立的<span style="color: black;">过程</span>。在结构方程中,这两步<span style="color: black;">同期</span>进行,即因子与题目之间的关系和因子与因子之间的关系<span style="color: black;">同期</span><span style="color: black;">思虑</span>。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(四)容许更大弹性的<span style="color: black;">测绘</span>模型</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">传统上,<span style="color: black;">咱们</span>只容许每一题目(指标)从属于单一因子,但结构方程分析容许更加<span style="color: black;">繁杂</span>的模型。例如,<span style="color: black;">咱们</span>用英语书写的数学试题,去<span style="color: black;">测绘</span>学生的数学能力,则测验得分(指标)既从属于数学因子,<span style="color: black;">亦</span>从属于英语因子(<span style="color: black;">由于</span>得分<span style="color: black;">亦</span>反映英语能力)。传统因子分析难以处理一个指标从属多个因子<span style="color: black;">或</span><span style="color: black;">思虑</span>高阶因子等有比较<span style="color: black;">繁杂</span>的从属关系的模型。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">(五)估计<span style="color: black;">全部</span>模型的拟合程度</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">在传统<span style="color: black;">路径分析</span>中,<span style="color: black;">咱们</span>只估计每一路径(变量间关系)的强弱。在结构方程分析中,除了<span style="color: black;">以上</span>参数的估计外,<span style="color: black;">咱们</span>还<span style="color: black;">能够</span>计算<span style="color: black;">区别</span>模型对同一个样本数据的整体拟合程度,从而判断哪一个模型更接近数据所呈现的关系。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">以上</span>内容是搜索告诉<span style="color: black;">咱们</span>关于结构方程模型的基本描述,比较重点的<span style="color: black;">地区</span><span style="color: black;">咱们</span>进行了标红,除此之外,<span style="color: black;">中间商</span>调节效应的分析在SEM下<span style="color: black;">亦</span>是很好分析的,并且随着普通多元回归的<span style="color: black;">乱用</span>,<span style="color: black;">通常</span>的统计模型<span style="color: black;">已然</span><span style="color: black;">不可</span>满足实证分析中需要推断因果关系的<span style="color: black;">需求</span>,<span style="color: black;">因此</span><span style="color: black;">此时</span>候就需要多种统计模型来进行分析,当然,SEM<span style="color: black;">亦</span>是有<span style="color: black;">必定</span>不足的?在之前<span style="color: black;">中间商</span>调节效应初探中说过,SEM的一个特点<span style="color: black;">便是</span>数据导向,<span style="color: black;">然则</span><span style="color: black;">实质</span>分析中,是需要理论来证明的,并且需要的样本量<span style="color: black;">亦</span>比<span style="color: black;">很强</span>,<span style="color: black;">针对</span>数据信效度的<span style="color: black;">需求</span>页比较高。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">为了让<span style="color: black;">大众</span><span style="color: black;">针对</span>SEM有比较系统的认识,<span style="color: black;">咱们</span>需要追根溯源来给<span style="color: black;">大众</span>介绍,<span style="color: black;">通常</span>的书籍中,将SEM分成两个部分,一个是结构模型,一个是<span style="color: black;">测绘</span>模型,当然最先要<span style="color: black;">处理</span>的是<span style="color: black;">测绘</span>模型,<span style="color: black;">倘若</span><span style="color: black;">测绘</span>模型存在问题,<span style="color: black;">那样</span>结构模型必然有巨大问题,<span style="color: black;">针对</span><span style="color: black;">详细</span>的这部分内容,<span style="color: black;">咱们</span>在后续会逐步给<span style="color: black;">大众</span>介绍。而今天为了让<span style="color: black;">大众</span><span style="color: black;">认识</span><span style="color: black;">测绘</span>模型,<span style="color: black;">咱们</span>给<span style="color: black;">大众</span>普及<span style="color: black;">有些</span>基本的知识,方便<span style="color: black;">大众</span>理解<span style="color: black;">测绘</span>模型的内容。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">古典<span style="color: black;">测绘</span>模型</span><span style="color: black;">:初中<span style="color: black;">理学</span>,<span style="color: black;">测绘</span>水的温度,需要三次<span style="color: black;">或</span><span style="color: black;">更加多</span>次数的<span style="color: black;">测绘</span><span style="color: black;">而后</span>取得平均值,这是自然科学<span style="color: black;">办法</span>,在社会科学中<span style="color: black;">咱们</span>怎么<span style="color: black;">测绘</span>呢?在大<span style="color: black;">都数</span><span style="color: black;">科研</span>中,<span style="color: black;">由于</span>成本问题,显然不可能多次<span style="color: black;">测绘</span>,且社会科学中<span style="color: black;">非常多</span>变量并不客观,<span style="color: black;">那样</span>这个<span style="color: black;">测绘</span>的过程中就会<span style="color: black;">显现</span>误差。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">下面引用上面<span style="color: black;">说到</span>书中的例子<span style="color: black;">来讲</span>明:</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/62icVzlxadTQuEGqibqvxUibzUpeUqg3TA6kL8oGsUJ5LM78etbjcpX1Gakr7k3au0FAF6HQMAYhLk2jiaSGmvz8CQ/640?wx_fmt=png&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">在<span style="color: black;">实质</span><span style="color: black;">测绘</span>中,真实值和观测值是会有<span style="color: black;">必定</span>误差的,而在普通的回归中,<span style="color: black;">咱们</span>默认得到的<span style="color: black;">测绘</span>值<span style="color: black;">便是</span><span style="color: black;">实质</span>值,这个假设有<span style="color: black;">必定</span><span style="color: black;">恰当</span>性,当然<span style="color: black;">亦</span>是有<span style="color: black;">必定</span>问题的</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">针对</span>古典测量模型<span style="color: black;">来讲</span>,公式很简单:X=q+w(被<span style="color: black;">叫作</span>作古典<span style="color: black;">测绘</span>模型)误差服从(0,1)正态分布。(1904,斯皮尔曼在<span style="color: black;">科研</span><span style="color: black;">关联</span>系数的时候发明的),<span style="color: black;">发掘</span>真实的<span style="color: black;">关联</span>比<span style="color: black;">测绘</span>的<span style="color: black;">关联</span>大,<span style="color: black;">因此</span><span style="color: black;">怎样</span><span style="color: black;">提高</span><span style="color: black;">测绘</span>的精度?</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">最简单的<span style="color: black;">处理</span><span style="color: black;">办法</span><span style="color: black;">便是</span><span style="color: black;">测绘</span>多个,原理是什么呢?</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">由于</span><span style="color: black;">咱们</span>在<span style="color: black;">测绘</span>的时候,指标与真实的数值之间都会有<span style="color: black;">必定</span>的差异,这个差异是随机的,正向负向,而多个<span style="color: black;">测绘</span>指标之间会<span style="color: black;">显现</span>正负项的相互抵消。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">除了古典<span style="color: black;">测绘</span>模型之外,同属<span style="color: black;">测绘</span>模型:与古典<span style="color: black;">测绘</span>模型有很大的相同,<span style="color: black;">然则</span>本质<span style="color: black;">区别</span>是在于<span style="color: black;">能够</span><span style="color: black;">准许</span><span style="color: black;">每一个</span>指标<span style="color: black;">表率</span>的比重不<span style="color: black;">同样</span>(权重)。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">而在结构方程模型中,<span style="color: black;">测绘</span>模型<span style="color: black;">便是</span>在进行模型拟合之前,<span style="color: black;">针对</span>得到数据的质量进行一个初步分析,当然这个分析<span style="color: black;">便是</span><span style="color: black;">经过</span>信效度来分,<span style="color: black;">针对</span>信效度的分析,SPSS做起来相对容易直观,stata的分析<span style="color: black;">咱们</span>在前面的分享中<span style="color: black;">亦</span>有介绍。而量表信效度的分析在SEM<span style="color: black;">亦</span><span style="color: black;">能够</span>完成,其核心思想与SPSS中并<span style="color: black;">无</span>什么<span style="color: black;">区别</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"> 简单的<span style="color: black;">来讲</span>,信度<span style="color: black;">便是</span>量表题目<span style="color: black;">测绘</span>的</span><span style="color: black;">稳定性,重复性</span><span style="color: black;">。<span style="color: black;">然则</span>从专业的<span style="color: black;">方向</span>说,信度被定义为一个<span style="color: black;">测绘</span>工具免于随机误差的影响程度。理解这个信度的概念还需要<span style="color: black;">测绘</span>的<span style="color: black;">关联</span>知识,<span style="color: black;">这里</span><span style="color: black;">弥补</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">回想<span style="color: black;">咱们</span>之前说的经典<span style="color: black;">测绘</span>模型,观测值(<span style="color: black;">测绘</span>分数)的<span style="color: black;">构成</span><span style="color: black;">是由于</span>真实值以及随机误差<span style="color: black;">构成</span>,信度指数<span style="color: black;">便是</span>真实值的方差除以观测值的<span style="color: black;">办法</span>,<span style="color: black;">然则</span>真实值<span style="color: black;">咱们</span>是不<span style="color: black;">晓得</span>的。然而真实的数据收集的过程中,误差不可能是仅仅是随机误差。<span style="color: black;">不外</span>再说下去就可能是<span style="color: black;">测绘</span>学<span style="color: black;">关联</span>的知识了,接下来<span style="color: black;">咱们</span>给大家介绍一下<span style="color: black;">关联</span>的信度。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">1. 复本信度,<span style="color: black;">区别</span>再测的<span style="color: black;">关联</span>系数</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">复本信度都是到<span style="color: black;">便是</span>用<span style="color: black;">区别</span>的工具测同一个构念,不需要是同一个对象,只需要是同一类适用人群,而得到的两份数据的<span style="color: black;">关联</span>系数就<span style="color: black;">能够</span>说明复本信度的高低。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">2. 重测信度(再测信度)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">两个<span style="color: black;">区别</span>的时间点,用同一工具<span style="color: black;">针对</span>同一<span style="color: black;">测绘</span>对象(这<span style="color: black;">通常</span><span style="color: black;">来讲</span>很难做到,<span style="color: black;">因此</span><span style="color: black;">运用</span>的很少),<span style="color: black;">一样</span><span style="color: black;">亦</span>是用<span style="color: black;">关联</span>系数去确定重测信度</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">3.内部一致性信度</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">a系数(<span style="color: black;">由于</span>那个字母<span style="color: black;">欠好</span>打),<span style="color: black;">评估</span>各指标之间的一致性(同质性),要明白内部一致性信度,必须先<span style="color: black;">晓得</span>量表长度与信度之间的关系。简单的说量表越长,信度越高,依据为斯皮尔曼-布朗公式,这<span style="color: black;">亦</span>是和信度的<span style="color: black;">测绘</span>目的相互匹配的。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">通常</span>都会报告a系数和<span style="color: black;">关联</span>系数,<span style="color: black;">通常</span>a系数0.7以上合格,而0.7以下不合格,0.8以上良好。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">在<span style="color: black;">通常</span>的分析中<span style="color: black;">咱们</span>仅仅需要关注a系数,<span style="color: black;">因此</span><span style="color: black;">这儿</span>有必要对a系数做出深入的<span style="color: black;">认识</span>。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">公式说明</span></strong><strong style="color: blue;"><span style="color: black;"> N/N-1 (</span></strong><strong style="color: blue;"><span style="color: black;">总方差—<span style="color: black;">每一个</span>题目方差之和</span></strong><strong style="color: blue;"><span style="color: black;">)/</span></strong><strong style="color: blue;"><span style="color: black;">总方差</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">A</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">B</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">C</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">VARIANCE</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">1</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">6</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">5</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">1</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">2</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">6</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">5</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">1</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">3</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">5</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">3</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">3</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">1.33</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">0</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">5</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">5</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">4</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">0.33</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">3.67</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">totly</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><strong style="color: blue;"><span style="color: black;">7</span></strong></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><img src="https://mmbiz.qpic.cn/mmbiz_png/62icVzlxadTQuEGqibqvxUibzUpeUqg3TA6rMboPLXplzRX4dLk0oSk2cgpDy9J0TL0YIgicxoLvveE0FT4icrEOibjA/640?wx_fmt=png&tp=webp&wxfrom=5&wx_lazy=1&wx_co=1" style="width: 50%; margin-bottom: 20px;"></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">上图引用自书中。a系数的一个特点是,随着题目数量的<span style="color: black;">增加</span>,系数值会增大。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">而<span style="color: black;">针对</span>效度</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">简单的说效度<span style="color: black;">便是</span><span style="color: black;">咱们</span><span style="color: black;">测绘</span>的东西<span style="color: black;">是不是</span>是<span style="color: black;">咱们</span>想<span style="color: black;">测绘</span>的构念。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">效度:<span style="color: black;">咱们</span>将问卷中的每一道题<span style="color: black;">作为</span>条目(项目)<span style="color: black;">或</span>是他所反映的一个指标,而<span style="color: black;">测绘</span>同一个构念的一组指标则被<span style="color: black;">作为</span>量表、量表是为了<span style="color: black;">测绘</span>某个构念而设计的,量表是否能真实反映<span style="color: black;">咱们</span>想测试的构念,这个<span style="color: black;">便是</span>量表的效度。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">关于效度检验:效度<span style="color: black;">重点</span>分四个方面:内容效度,结构效度,效标效度,构念效度。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">1.内容效度</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">1.<span style="color: black;">测绘</span>的内容<span style="color: black;">是不是</span>充分反映了这个构念(如,一个构念有N个指标,而你只用了N-n个指标,其中N>n)2.指标的<span style="color: black;">选择</span><span style="color: black;">是不是</span>有<span style="color: black;">表率</span>性(如,一个构念有N个指标,<span style="color: black;">然则</span>你用的指标中有与构念不符合的,<span style="color: black;">例如</span>满意度的量表变<span style="color: black;">成为了</span>幸福感的,<span style="color: black;">或</span>工作满意度的用生活满意度<span style="color: black;">测绘</span>)3.问卷形式需要符合当地文化。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">盲人摸象的故事<span style="color: black;">便是</span>说明内容效度不足,<span style="color: black;">测绘</span>需要准确覆盖构念的<span style="color: black;">每一个</span>部分(<span style="color: black;">重点</span><span style="color: black;">办法</span>有<span style="color: black;">规律</span>判断法,专家分析法和实证<span style="color: black;">科研</span>法)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">2.(内部)结构效度</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">测绘</span>数据得到的构念的结构<span style="color: black;">是不是</span>于<span style="color: black;">咱们</span>对构念的预期一致。(因子分析是<span style="color: black;">处理</span>结构效度的很好的<span style="color: black;">办法</span>,因子分析<span style="color: black;">能够</span>帮我<span style="color: black;">认识</span>一个<span style="color: black;">测绘</span>项目<span style="color: black;">背面</span>有多少个因子)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">内容结构效度是<span style="color: black;">咱们</span><span style="color: black;">通常</span>论文中量表效度所指。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">探索性因子分析在<span style="color: black;">研发</span>量表中有很大的<span style="color: black;">功效</span>,当构念<span style="color: black;">不足</span>清楚的时候,<span style="color: black;">咱们</span>需要<span style="color: black;">经过</span>数据去探索这个概念的维度。而验证性是在<span style="color: black;">已然</span><span style="color: black;">晓得</span>维度的<span style="color: black;">状况</span>下去验证,<span style="color: black;">通常</span>用来检验别人<span style="color: black;">已然</span><span style="color: black;">研发</span>和使用的量表。相对<span style="color: black;">来讲</span>,探索性因子分析的<span style="color: black;">需求</span>比验证性更加苛刻,<span style="color: black;">通常</span>探索性<span style="color: black;">不外</span>的话<span style="color: black;">亦</span>不需要直接删除题目。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">3.基于该<span style="color: black;">测绘</span>与其他<span style="color: black;">测绘</span>关系之间的效度(效标效度,<span style="color: black;">规律</span>关系网)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">效标效度(又<span style="color: black;">叫作</span>效标<span style="color: black;">相关</span>效度)举例<span style="color: black;">测绘</span>了A和B,并且理论<span style="color: black;">已然</span>告诉<span style="color: black;">咱们</span>AB<span style="color: black;">最少</span>有比较强的<span style="color: black;">关联</span>性(不<span style="color: black;">必定</span>是因果),<span style="color: black;">然则</span><span style="color: black;">发掘</span>AB不<span style="color: black;">关联</span>,<span style="color: black;">那样</span><span style="color: black;">通常</span><span style="color: black;">便是</span>AB的<span style="color: black;">测绘</span>出了问题。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">规律</span>关系网,简单的说<span style="color: black;">便是</span>你的数据验证了理论<span style="color: black;">或</span>模型,<span style="color: black;">那样</span>你的数据就应该是有效度的。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">4.聚合效度和区别效度(被误解最多的效度)</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">科研</span>中的因子分析并<span style="color: black;">不可</span>说明<span style="color: black;">咱们</span>的问卷<span style="color: black;">拥有</span>聚合效度<span style="color: black;">或</span>区分效度。<span style="color: black;">由于</span>因子分析只能说明量表<span style="color: black;">能够</span>把几个概念区<span style="color: black;">掰开</span>,<span style="color: black;">然则</span>并<span style="color: black;">不可</span>说明有效度。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">聚合效度的意思是一个构念的<span style="color: black;">测绘</span>与<span style="color: black;">表率</span><span style="color: black;">一样</span><span style="color: black;">测绘</span><span style="color: black;">关联</span>性很大,而区别效度是一个构念的<span style="color: black;">测绘</span>和其他构念<span style="color: black;">关联</span>很小。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">但在结构方程模型中,聚合效度(construct validity)因子载荷大于0.71说明题目有很大的聚合,组合效度(Hair,2006),<span style="color: black;">为何</span>呢?<span style="color: black;">由于</span>0.71x0.71>0.5,解释超过百分之五十的方差变异,而在社会科学中0.55以上<span style="color: black;">便是</span><span style="color: black;">能够</span>接受的范围。因子载荷在下一节中讲</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">多质多法检验量表的聚合和区别效度</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">在<span style="color: black;">研发</span>量表的时候极为重要,举例,<span style="color: black;">例如</span>一个构念A的量表(来自美国社会科学<span style="color: black;">行业</span>),<span style="color: black;">咱们</span>觉得A的<span style="color: black;">测绘</span>的不准确(<span style="color: black;">亦</span><span style="color: black;">便是</span>通俗的不符合文化环境),<span style="color: black;">那样</span><span style="color: black;">咱们</span>需要<span style="color: black;">研发</span>一个A+,<span style="color: black;">表率</span>有中国<span style="color: black;">特殊</span>的问卷,此时<span style="color: black;">咱们</span>需要引入B(理论上<span style="color: black;">区别</span><span style="color: black;">然则</span>很相信的概念),此时,将三个问卷放入一个数据中,<span style="color: black;">此时</span>候<span style="color: black;">倘若</span>A+与A<span style="color: black;">关联</span>系数<span style="color: black;">很强</span>,且A+与B<span style="color: black;">关联</span>系数较小,则说明新的量表A+有区分和聚合效度,而这个区分和聚合效度是针<span style="color: black;">针对</span><span style="color: black;">区别</span>概念的区分,以及<span style="color: black;">一样</span>概念的聚合。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">此<span style="color: black;">办法</span>是campbell以及Fiske在1959年提出的,多质多法矩阵(MTMM),需要自评和他<span style="color: black;">评估</span>以及两个以上的构念进行<span style="color: black;">关联</span>矩阵的输出,多于一个的<span style="color: black;">办法</span><span style="color: black;">测绘</span>多于一个的质。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;"><span style="color: black;">针对</span>NTMM<span style="color: black;">倘若</span>想有深入<span style="color: black;">认识</span>的话,<span style="color: black;">大众</span><span style="color: black;">能够</span><span style="color: black;">查询</span><span style="color: black;">关联</span>书籍。</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">接下来是效度的干货内容</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">而在<span style="color: black;">通常</span>的论文中,<span style="color: black;">咱们</span>是<span style="color: black;">没法</span>满足这么多效度的检验,<span style="color: black;">因此</span>,<span style="color: black;">通常</span><span style="color: black;">咱们</span>采取的是</span></p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">建构效度</p>建构效度分为两种:收敛效度与区别效度。收敛效度是<span style="color: black;">指的是</span>相同概念里的项目,彼此之间<span style="color: black;">关联</span>度高。区别效度<span style="color: black;">指的是</span><span style="color: black;">区别</span>概念里的项目,彼此<span style="color: black;">关联</span>度低。<span style="color: black;">简单<span style="color: black;">来讲</span><span style="color: black;">便是</span><span style="color: black;">区别</span>构念(潜变量)之间的<span style="color: black;">关联</span>度要<span style="color: black;">少于</span>构念内部的<span style="color: black;">关联</span>度。这个<span style="color: black;">原由</span>很好理解,<span style="color: black;">咱们</span><span style="color: black;">科研</span><span style="color: black;">区别</span>构念之间的关系,当然是<span style="color: black;">期盼</span><span style="color: black;">咱们</span>的自变量与因变量有很强的<span style="color: black;">关联</span>性,最好还是<span style="color: black;">掌控</span><span style="color: black;">有些</span>变量<span style="color: black;">状况</span>下的净<span style="color: black;">关联</span>性,<span style="color: black;">这般</span><span style="color: black;">才可</span>说明<span style="color: black;">咱们</span>关注的点是正确的,我们<span style="color: black;">能够</span>简单理解成,<span style="color: black;">期盼</span><span style="color: black;">区别</span>构念之间的<span style="color: black;">关联</span>性要很强,<span style="color: black;">这般</span>得到的回归系数<span style="color: black;">亦</span>会越大越<span style="color: black;">明显</span>,<span style="color: black;">然则</span>这<span style="color: black;">亦</span>是需要有一个限度的,而怎么确定这个限度的,为了方便,在<span style="color: black;">科研</span>中,<span style="color: black;">咱们</span>关注的是构念之间的<span style="color: black;">关联</span>和构念内部的<span style="color: black;">关联</span>的关系,<span style="color: black;">因此</span>内部的关系<span style="color: black;">必定</span>要强于<span style="color: black;">外边</span>的关系。这<span style="color: black;">便是</span><span style="color: black;">表现</span>区别,而收敛<span style="color: black;">便是</span>自己的构念题目内,要有<span style="color: black;">必定</span><span style="color: black;">关联</span>性,而这个<span style="color: black;">关联</span>性就用AVE进行<span style="color: black;">测绘</span>。</span><span style="color: black;">
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">对<span style="color: black;">测绘</span>模型进行验证性因子分析来检验<span style="color: black;">重点</span>潜变量的收敛效度(convergent validity)和判别效度(discriminant validity)(Anderson and Gerbing,1 988)128]。</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">1. 收敛效度检验。对三个量表即三个<span style="color: black;">测绘</span>模型进行收敛效度检验时,<span style="color: black;">第1</span>步考察每一个潜变量的标准化因子载荷系数,载荷值应>0.5,这<span style="color: black;">寓意</span>着问</p>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;">项与其潜变量之间的<span style="color: black;">一起</span>方差大于问项与误差方差之间的<span style="color: black;">一起</span>方差,都是<span style="color: black;">明显</span>的;第二步考察AVE值。AVE值应>0.5,这<span style="color: black;">寓意</span>着每一个因子所提取的可解释50%以上的方差(Fomell和Larcker,1981)</p>2. 判别效度检验。<span style="color: black;">针对</span>各维度间<span style="color: black;">是不是</span>存在足够的区分效度,<span style="color: black;">一般</span>采用比较各维度间完全标准化<span style="color: black;">关联</span>系数与所<span style="color: black;">触及</span>各维度<span style="color: black;">自己</span>AVE的平方根值<span style="color: black;">体积</span>,当前者<span style="color: black;">少于</span>后者,则<span style="color: black;">显示</span>各维度间存在足够的区分效度,反之,则区分效度<span style="color: black;">不足</span>。(Fomell&Larcker,1981)
</span>
<p style="font-size: 16px; color: black; line-height: 40px; text-align: left; margin-bottom: 15px;"><span style="color: black;">今天的分享<span style="color: black;">重点</span><span style="color: black;">便是</span><span style="color: black;">测绘</span>模型需要<span style="color: black;">认识</span>的前提知识<span style="color: black;">亦</span>是<span style="color: black;">针对</span>信效度的一个回顾和加深</span></p>
你的见解真是独到,让我受益良多。 论坛外链网http://www.fok120.com/
页:
[1]