Stata 17中如何在汇总统计表格中添加变量正态性检验的P值？

Hey there! Great question—adding normality test p-values to your summary statistics table in Stata 17 is totally achievable, and I’ve got a few practical methods to help you pull this off smoothly.

This is my go-to approach because it lets you generate clean, customizable tables that include both standard summary stats and normality test p-values. First, you’ll need the estout package (it’s free from Stata’s SSC repository):

* Install estout if you haven't already
ssc install estout

* Define your list of variables (replace with your actual variable names)
local analysis_vars age income score

* Clear existing estimates to start fresh
estimates clear

* Loop through each variable to collect stats and normality p-values
foreach var of local analysis_vars {
    * Grab standard summary statistics
    estpost summarize `var'
    
    * Run Shapiro-Wilk normality test and save the p-value
    swilk `var'
    local sw_p = r(p)
    estadd scalar shapiro_wilk_p = `sw_p'
    
    * Optional: Add Kolmogorov-Smirnov test p-value (uses sample mean/sd as distribution params)
    ksmirnov `var' = normal(mean(`var'), sd(`var'))
    local ks_p = r(p)
    estadd scalar kolmogorov_smirnov_p = `ks_p'
    
    * Store the results for this variable
    estimates store `var'
}

* Generate the final table with your desired stats
esttab *, cells("mean sd min max shapiro_wilk_p kolmogorov_smirnov_p") ///
        label ///
        title("Summary Statistics with Normality Test P-Values") ///
        format(%9.4f) ///
        replace

What this does: #

• The loop processes each variable individually, collecting mean, standard deviation, min/max, and p-values from two common normality tests.
• estadd scalar attaches the p-values to the summary stats so they show up in the final table.
• The esttab command lets you tweak formatting (like decimal places, labels, and title) to fit your needs.

If you prefer not to install estout, you can manually compile stats and p-values into a matrix for display:

* Define your variables
local analysis_vars age income score

* Get basic summary stats and save them to a matrix
tabstat `analysis_vars', stats(mean sd min max) save
matrix basic_stats = r(StatTotal)

* Create a matrix to hold normality p-values (2 rows for SW and KS tests)
matrix norm_pvals = J(2, colsof(basic_stats), .)

* Fill the p-value matrix
forvalues i = 1/colsof(basic_stats) {
    local current_var = colnames(basic_stats)[`i']
    
    * Shapiro-Wilk test
    swilk `current_var'
    matrix norm_pvals[1,`i'] = r(p)
    
    * Kolmogorov-Smirnov test
    ksmirnov `current_var' = normal(mean(`current_var'), sd(`current_var'))
    matrix norm_pvals[2,`i'] = r(p)
}

* Combine the basic stats and p-value matrices
matrix full_stats = basic_stats \ norm_pvals

* Add row/column labels for clarity
matrix colnames full_stats = `:colnames(basic_stats)'
matrix rownames full_stats = "Mean" "Standard Deviation" "Minimum" "Maximum" "Shapiro-Wilk P" "Kolmogorov-Smirnov P"

* Display the final matrix
matrix list full_stats, format(%9.4f)

If you just want to view stats and p-values side-by-side without a formal table, a simple loop will do:

local analysis_vars age income score

foreach var of local analysis_vars {
    di "====================================="
    di "Summary Statistics for: `var'"
    di "====================================="
    summarize `var'
    di ""
    swilk `var'
    di "Shapiro-Wilk Normality Test P-Value: " string(r(p), "%9.4f")
    ksmirnov `var' = normal(mean(`var'), sd(`var'))
    di "Kolmogorov-Smirnov Normality Test P-Value: " string(r(p), "%9.4f")
    di ""
}

A Quick Note on Normality Tests: #

• Shapiro-Wilk: Best suited for small to medium-sized samples (typically n ≤ 2000; Stata may restrict it for larger samples).
• Kolmogorov-Smirnov: Works for larger samples, but note that we’re using the sample mean and standard deviation to define the normal distribution here (this is the "Lilliefors" variant in practice).

内容的提问来源于stack exchange，提问作者Patudb