1、排序sort cases by varname (a/b).数据合并add files/file=*/rename old varname=new varname/drop varname.数据转换Computecompute target varname=expression.算术运算:+,-,*,/,*(幂)算术函数:sqrt, rnd(四舍五入), trunc(取整)统计函数:mean, sum, sd, max, min 缺失值函数:SYSMIS,MISSING,NMISS,NVALID,VALUE不算缺失值时间函数: CTIME.DAYS,$JDATEIfif () target v
2、arname=expression.Recoderecode varname (old value /else/lowest through value/value through value/value through high=new value) into newvarname.Missing values:user-defined 也会被重编码,因此应小心user-defined. user-defined 不包括在范围内Split fileSort cases by varname .split file layered by varname .split file off. Fli
3、pflip variables = varname/all.Rankrank variable = varname.数据分析描述性统计FrequenciesFrequencies varname/histogram NORMAL /barchart (freq/percent)/piechart.DescriptiveDescriptive varname/statistics=sum mean min max (集中趋势) RANGE stddev SEMEAN VARIANC(离中趋势) skewness偏度 SESKEW KURTOSIS峰度 SEKURT形状测量/varname (zn
4、ewvarname).Exploreexamine varname/plot BOXPLOT STEAMLEAF HISTOGRAM(不带有正态曲线) NPPLOT证明正态/ STATISTICS DESCRIPTIVES EXTREME/MESTIMATORS修正均值.Crosstabs/tables=row varname BY column varname/Statistic=chisq默认 corr KAPPA评定者间一致性系数PHI CC修正/CELLS= .相关分析Graphscatterplot=varname with varname.Personcorrelations/va
5、riables=varlist/ MISSING=PAIRWISE/LISTWISE数据足够时更稳定.Spearman/kendall(顺序变量)nonpar corr/print=spearman(默认)/kendall有重复/both.差异性检验单样本检验t-test testval=value/variables=varname.独立样本检验t-test groups=varname(value1 value2)相关样本检验t-test pairs=varname with varname.曼惠特尼U检验npar tests/ m-w=varname with value1 value2
6、.维克尔松检验npar test/ wilcoxon=varname with varname.方差分析前提假设:独立、等距、正态、同质数据要求:多元正态,线性(散点图)Oneway anova单一自变量平-单一因变量oneway varlist by varname/statistics descriptives homogeneity/contrast valuelist/posthoc=LSD TURKEY敏感BONFERRONI Scheffe(保守) snk DUNNETT.Unianova多个自变量-单一因变量unianova dependent varname BY factor
7、 varlist/posthoc varlist=lsd snk turkey/plot=profile(varname/varname*varname)/desigh=factor varlist.manovadependent varname BY factor varlist/DESIGN=FACNAME1 within FACNAME2(FACNUM1) FACNAME1 within FACNAME2(FACNUM2)Multivariate多个自变量-多个因变量glmdependent varlist BY factor varlist/desigh=factor varlist/
8、PRINT = HOMOGENEITY同质性前提 dependent varlist BY factor varlistREPEATED MEASURE包含组内自变量-多个因变量wsfactor varlist BY factor varlist/WSFACTORs wsfacname wsfacnum/wsdisigh wsfaclist/desigh=varlist/emmeasn = tables(varname) compare adj(lsd)/plot=profile(varname/varname*varname/varname*wsfacname)manova wsfactor
9、 varlist BY facname (facnum)/WSFACTORS wsfacname (wsfacnum)/WSDESIGN =WSFACNAME1 within WSFACNAME2(WSFACNUM1) WSFACNAME1 within WSFACNAME2(WSFACNUM2)/WSDESIGN=WSFACNAME/DESIGN=MWITHIN FACNAME (FACNUM1) MWITHIN FACNAME (FACNUM2) /DESIGN=FACNAME/WSDESIGN=MWITHIN WSFACNAME (WSFACNUM1) MWITHIN WSFACNAME
10、 (WSFACNUM2).回归分析等距/等比/顺序 线性 非共线性 残差正态,同质,线性cases : variables=10:1;被试数目 100;无Outliers;无multicollinearityIGRAPH /VIEWNAME=Scatterplot /X1 = VAR(before) TYPE = SCALE /Y = VAR(after) TYPE = SCALE /COORDINATE = VERTICAL /FITLINE METHOD = REGRESSION LINEAR LINE = TOTAL SPIKE=OFF /X1LENGTH=3.0 /YLENGTH=3.
11、0 /X2LENGTH=3.0/CHARTLOOK=NONE/SCATTER COINCIDENT = NONE.EXE.REGRESSION/DEPENDENT DEPENDENT VARNAME/METHOD=STEPWISE/ENTER FACTOR VARLIST/STATISTICS COEFF OUTS R ANOVA COLLIN检验multicollinearity/CASEWISE检验outlier .聚类分析CLUSTER VARLIST/PRINT SCHEDULE CLUSTER(CLUNUM)/PLOT DENDRGRAM VICICLE主成分分析多元正态 线性 等距
12、顺序 相关- variables=5:200;正态不是必须,如果正态分布,解决会更好;线性:如果非线性,应考虑转换变量后再作因素分析;无Outliers;在主成分分析中,multicollinearity不是问题,在主因素分析中,不能有 multicollinearity相关矩阵FACTOR /MATRIX=IN(CORR=*) /method=correlation /format=sort blank(0.40) /ROTATION VARIMAX /PLOT=EIGEN rotation.一般主成分分析 /VARIABLES VARLIST /PLOT EIGEN ROTATION /R
13、OTATION VARIMAX/METHOD=CORRELATION /PRINT KMO 相关和共线性检验.数据报告报告Descriptives峭度为正表示总体分布的峰态较标准正态更陡; 反之更缓.斜度为正表示样本值比拟集中于均值的左边; 斜度为负表示样本值比拟集中于均值的右边去掉5%的均值5% trimmed mean 四分位距interquartile range读Respondents Stem-and-Leaf Plot读Respondents BoxplotOutlier:从矩形框始,在1.5 倍箱距的点之外Extreme:从矩形框始,在3倍箱距的点之外 1.5 倍箱距的点之外的O
14、utlier需要给予注意,如果是多于3个点位置很近,多数情况考虑保存。3 倍箱距的点之外的Extreme value需要给予特别注意,如果是孤立的点,多数情况考虑可以作为缺失值计算一次,作为有效值计算一次。读Normal Q-Q Plot of Respondents X轴:实得分数 Y轴:Z分数的期望值 如果呈线性说明正态分布报告Respondents A*B Crosstabulation单位格 standardized residual 2.0引起chi-square 增加或显著报告Chi-Square Tests报告自由度 df = (r-1) * (c-1)E 的最小值 5, c 可
15、能不准确Chi Square 统计量不是描述相关的良好指标。因为它随样本量变化而变化。两变量间各种类型的相关都会产生一样的Chi Square 值。报告Graph散点图提示的outlier 需要特别处理散点图也帮助我们找出multivariate outlier,关系的异常值报告Correlations-Correlations和Nonparametric Correlations - Correlation对数据的相关分析显示,a与b有显著相关,r(df)=,p.001,双尾报告One-Sample Test报告Independent Samples Test如果Levenes Test f
16、or Equality of Variances不显著,报告Equal variances assumed 的Welchst test结果。s Test for Equality of Variances显著,报告Equal variances not assumed 的t test 结果。a的均值(M =, SD =) 与bM =, SD =)有显著差异 。t(df) =, p = 0.05。报告Paired Samples Test曼惠特尼U检验独立样本报告Test Statistics维克尔松检验相关样本One way anova看Descriptives报告ANOVA报告Multipl
17、e Comparisonsa的均值(M =, SD =) 、b的均值M =, SD =)和c的均值M =, SD =)有显著差异 。事后检验显示,a与b有显著差异,a-b=,p=0.05; c与b有显著差异,c-b=,p= a与c有显著差异,a-c=,p=Unianova主效应和交互作用报告Tests of Between-Subjects Effects自变量只有两组时,报告Hotellings Trace;自变量大于两组时,报告Wilks Lambda;方差齐性的统计前提被违反时,报告Pillais Trace 。如果是不平衡处理,事后检验报告Estimated Marginal Mean
18、s- Pairwise ComparisonsOption如果是平衡处理报告,事后检验报告Pairwise ComparisonsPosthoc交互作用图报告Profile Plots*因素方差分析结果显示,a与b交互作用显著,F=,p0.05,交互作用图如图1。简单效应结果显示,a1在b1M=,SD=和b2M=,SD=上有显著差异,F=,p0.05;a2在b1M=,SD=和b2M=,SD=上有显著差异,F=,pa的主效应显著,F=,pb的主效应显著,F=,pMultivariate报告Boxs Test of Equality of Covariance Matrices 简单效应结果报告M
19、anovaREPEATED MEASURE如果Mauchly test of sphericity检验不显著,交互作用、组内主效应结果报告Tests of Within-Subjects Effects如果Mauchly test of sphericity检验显著,交互作用、组内主效应结果报告Multivariate Tests 组建主效应报告Tests of Between-Subjects Effects组内变量事后检验结果报告Estimated Marginal Means-Pairwise Comparisons如果是不平衡处理,组间变量事后检验报告Estimated Margina
20、l Means- Pairwise ComparisonsOption如果是平衡处理报告,组间变量事后检验报告Pairwise ComparisonsPosthoc重复测量方差分析结果显示,a与b交互作用显著,F=,p 75% 很好; 50-75% 不错;25-50% 一般; 25% 不够 模型总效应报告ANOVABeta、b结果报告Coefficients以a,b,c为自变量,用逐步回归方法对d进行分析。回归的拟合度良好,R2adj=,模型的总效应F()=,p0.005。当其他变量恒定时,a与d有正相关,beta=0.653,其效应边缘显著,t()=,p0.6R的特征值和奉献率报告Total Variance Explained碎石图报告Scree plot主成分载荷报告Component Matrix因子旋转结果报告Rotated Component MatrixR的特征值和奉献率如表1。根据Kaiser准那么,取特征大于1的a为主成分,其积累奉献率到达c%,根本包括了全部指标所包括的信息。取前n个特征值,计算主成分载荷。由主成分分析的结论,对因子载荷阵实行方差最大旋转,可将m个指标分为n类,结果如表2:
copyright@ 2008-2022 冰豆网网站版权所有
经营许可证编号:鄂ICP备2022015515号-1