Calculating P-Values | Synth Runner | Example

[Konstantin Fischer]

Hello everyone,

I really need help with the understanding on how synth_runner calculates the Pvals / Pvals_std. Synth_runner (http://econweb.umd.edu/~galiani/files/synth_runner.pdf) gives me out the following results (post-treatment period). The data includes two treated units (treated on the same year) and eight un-treated units (placebo on the same year).

The results for the first post-treatment year:

Estimates: .0015411 [average treatment-effect of the two treated units]
PVals: .6428571
Pvals_std: .452381

The Stdv. for the two treated units is 0.003642265 (S) or 0.00257547 (N). The p-value is calculated as written on page 4 till 5 under 2.2. / 2.3 in the document.

When I run synth on all untreated units [excl. treated from the data-set] than I get for the eight institutions following averaged results for the same year:

Estimates: .000303053 [average of eight untreated units]

On an individual level I have the following estimates for the same year and unit:

[1] -0.00509145 | [2] 0.0019493 | [3] -0.00446586 | [4] 0.00220556 | [5] 0.00629673 |
[6] -0.00248195 | [7] -0.00384403 | [8] 0.00785612

The Stdv. on these estimates (un-treated units) is equal to 0.004691753 (N) or 0.005015695 (S).

When I calculate it by hand than my results for the Pvals / Pvals_std. are different.

Is there anyone that can help me with this issue, please?

Hope someone is able to do me a favor…

Thank you,

Konstantin

[Alexis Rodas]

Hi Konstantin,

I have the same problem too. But I have a question and I hope you can help me. How did you get the manual p-values without using the synth_runner command? If you could share your code, it is that I can not make a program to obtain the manual p-waltz.

Thank you very much for your help.

Alexis Rodas