非正态的样本這概念沒聽過。樣本應該與总体分布類似。如果抽樣方法有問題，以致样本與总体分布不同，這樣的檢驗沒有意義。
對总体非正态，可以考慮先將檢測值作轉化（正态化）。t-tets四乎也沒有用中位数的選項。

t检验检验的是样本均值与总体均值是否有显著性差异。

我用spss的单样本Wilcoxon符号秩和检验，检验样本的中位数和某个常量是否相等。

与用stata 的signtest命令，signtest 变量1=常数，得出的结果不一样。

请问您知道应该用哪种方法吗？

有没有大神知道？急，在线等

https://en.wikipedia.org/wiki/Median_test
参考看看

Stata 列出下列方法
http://www.stata.com/features/nonparametric-methods/

Nonparametric tests
Sign test
Wilcoxon signed-rank test
dialog box
Wilcoxon rank-sum (Mann–Whitney) test
Equality of medians
Kruskal–Wallis test
Kolmogorov–Smirnov test
Binomial probability test
Chi-squared tests for association
Tests for trend
One- and two-sample tests of proportions
Test for random order
Equality of survivor functions
Equality of ROC areas
Equality of ROC area against a standard ROC curve
CUSUM
Interrater agreement
Quantile regression
Median regression
Least absolute deviations regression
User-specified quantiles
Koenker and Bassett and Huber sandwich standard errors with many choices of bandwidth methods and kernels
Simultaneous quantile regression
Interquartile range regression
Nonparametric correlation coefficients
Spearman’s rank order
Kendall’s rank order
Survival analysis
Kaplan–Meier curves
Nelson–Aalen curves
Logrank and other tests of equality
Treatment effects
Nearest-neighbor matching
Propensity-score matching
Resampling and simulations
Bootstrap sampling and estimation
Sampling with replacement
Monte Carlo simulations
Smoothing
Univariate kernel density estimation
Lowess smoothing
Kernel-weighted local polynomial smoothing
Robust nonlinear smoother
ROC analysis
Adjust for covariates
Plot ROC curves  
Centiles and confidence intervals
Estimate percentiles
Exact confidence intervals
Optionally assume normality

你得知道你要測什麼