svy leads to insufficient observations

[Sven Johnsson]

I'm using the IPUMS provided CPS data (ASEC) and am running a wage regression. Basically, I do

svyset [iw=wtsupp], sdrweight(repwt1-repwt160) vce(sdr)
replace incwage = . if incwage == 9999999
replace incwage = . if incwage == 9999998
gen wage_log = log(incwage)
drop if year < 2000

gen food = ind1990 == 641
gen food_year = food * year

svy: reg wage_log I.food_year

which is basically regressing log wages onto a particulary industry code by year. However, I get the following output:

SDR replications (160)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 50
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 100
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 150
xxxxxxxxxx
insufficient observations to compute bootstrap standard errors
no results will be saved

What should I do?

[Sven Johnsson]

push

[Friedrich Huebler]

push

Please see "Why did my question not get answered?" and "Bumping".

[Marcos Almeida]

Friedrich gave excellent advice.

You may also wish to present further information on your data, such as using the - svydes - command. Please read the FAQ. There, you will find an advice on how to show "exactly" what your typed. That will help a lot and surely increase the odds of getting an insightful reply.

[Sven Johnsson]

And I was following the "advice" to the letter when I bumped a 3-days-nonresponsive thread once, after having had provided all the information I thought there was to it. As for "showing exactly what you typed", I can't find any reference to that in the FAQ.

Here goes - svydes- :

Survey: Describing stage 1 sampling units

      iweight: wtsupp
          VCE: sdr
          MSE: off
    sdrweight: repwt1 repwt2 repwt3 repwt4 repwt5 repwt6 repwt7 repwt8 repwt9 repwt10 repwt11 repwt12 repwt13 repwt14 repwt15 repwt16 repwt17
               repwt18 repwt19 repwt20 repwt21 repwt22 repwt23 repwt24 repwt25 repwt26 repwt27 repwt28 repwt29 repwt30 repwt31 repwt32 repwt33
               repwt34 repwt35 repwt36 repwt37 repwt38 repwt39 repwt40 repwt41 repwt42 repwt43 repwt44 repwt45 repwt46 repwt47 repwt48 repwt49
               repwt50 repwt51 repwt52 repwt53 repwt54 repwt55 repwt56 repwt57 repwt58 repwt59 repwt60 repwt61 repwt62 repwt63 repwt64 repwt65
               repwt66 repwt67 repwt68 repwt69 repwt70 repwt71 repwt72 repwt73 repwt74 repwt75 repwt76 repwt77 repwt78 repwt79 repwt80 repwt81
               repwt82 repwt83 repwt84 repwt85 repwt86 repwt87 repwt88 repwt89 repwt90 repwt91 repwt92 repwt93 repwt94 repwt95 repwt96 repwt97
               repwt98 repwt99 repwt100 repwt101 repwt102 repwt103 repwt104 repwt105 repwt106 repwt107 repwt108 repwt109 repwt110 repwt111
               repwt112 repwt113 repwt114 repwt115 repwt116 repwt117 repwt118 repwt119 repwt120 repwt121 repwt122 repwt123 repwt124 repwt125
               repwt126 repwt127 repwt128 repwt129 repwt130 repwt131 repwt132 repwt133 repwt134 repwt135 repwt136 repwt137 repwt138 repwt139
               repwt140 repwt141 repwt142 repwt143 repwt144 repwt145 repwt146 repwt147 repwt148 repwt149 repwt150 repwt151 repwt152 repwt153
               repwt154 repwt155 repwt156 repwt157 repwt158 repwt159 repwt160
  Single unit: missing
     Strata 1: <one>
         SU 1: <observations>
        FPC 1: <zero>

                                      #Obs per Unit
                              ----------------------------
Stratum    #Units     #Obs      min       mean      max  
--------  --------  --------  --------  --------  --------
       1 8,677,348 8,677,348         1       1.0         1
--------  --------  --------  --------  --------  --------
       1 8,677,348 8,677,348         1       1.0         1

[Marcos Almeida]

Sven Johnsson And I was following the "advice" to the letter when I bumped a 3-days-nonresponsive thread once, after having had provided all the information I thought there was to it. As for "showing exactly what you typed", I can't find any reference to that in the FAQ.

You may find this information precisely in the FAQ, as you may wish to read here.

Below, the excerpt, exactly with the advice previously mentioned.

12.1 What to say about your commands and your problem

Say exactly what you typed and exactly what Stata typed (or did) in response. N.B. exactly!

[Sven Johnsson]

Well, that's all in the original post (so I expected something beyond that which I was missing). For completeness, here are is the log for the earlier commands which I deemed irrelevant:

. replace incwage = . if incwage == 9999999
(2,034,734 real changes made, 2,034,734 to missing)

. replace incwage = . if incwage == 9999998
(1,628 real changes made, 1,628 to missing)

. gen wage_log = log(incwage)
(4,545,062 missing values generated)

. drop if year < 2000
(5,235,427 observations deleted)

. 
. gen food = ind1990 == 641

. gen food_year = food * year

. 
. svy: reg wage_log I.food_year
(running regress on estimation sample)

SDR replications (160)
----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5 
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx    50
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx   100
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx   150
xxxxxxxxxx
insufficient observations to compute bootstrap standard errors
no results will be saved
r(2000);

[Friedrich Huebler]

If you don't get an answer here you can contact Stata tech support.

[Marcos Almeida]

Browsing on the Web, we see there are similiar error messages when using the bootstrap method, and they are not necessarily related to svy prefix.

Actually, they may well happen on account of the bootstrapping replication approach itself.

These are the bad news.


The good news are:

First, under large samples, the bootstrap method will provide similar results, if compared to jackknife or balanced repeated replication methods.

Second, shall you select the jackknife approach, for example, it is quite easy to perform this method in Stata under svy prefix.