Issues displaying the exact results of panel FMOLS using the command xtcointreg

[Mir Adnan Ali]

Hello,
I'm using the command xtcointreg for estimating panel FMOLS. I used the following code:

xtcointreg ferti rpgdp hcfer women flfpr lgdpc agedp

Here are the results:

Method of estimation: FMOLS
Number

Cave[6,2]
         beta  t-stat
rpgdp   -0.01  -10.85
hcfer    0.62   63.31
women    0.00    0.75
flfpr   -0.00    6.89
lgdpc   -0.34  -12.87
agedp    0.01   10.68

Now, I'm building a table containing these results.

I used the following code:

collect clear
table () (command result), command(xtcointreg ferti rpgdp hcfer women flfpr lgdpc agedp)
collect label list result, all

table () (command result), command(_r_b _r_se _r_z _r_p: xtcointreg ferti rpgdp hcfer women flfpr lgdpc agedp) nformat(%5.4f  _r_b _r_se _r_z _r_p )
collect layout
collect dims
collect levelsof result
collect label list result, all

collect label levels result _r_se "Standard error", modify
collect label levels result _r_z "t-statistic", modify
collect label levels result _r_p "Probability", modify

collect label list command, all

collect label levels command 1 "Fully modified OLS estimation results", modify
collect stars _r_p 0.01 "***" 0.05 "**" 0.1 "*", attach(_r_p)

collect label levels colname rpgdp "RPGDP" hcfer "HCFER" women "WOMEN" flfpr "FLFPR" lgdpc "LGDPC" agedp "AGEDP" _cons "Constant", modify

collect preview

The preview of the table looks like this:

. collect preview

--------------------------------------------------------------------
         |            Fully modified OLS estimation results         
         |  Coefficient   Standard error   t-statistic   Probability
---------+----------------------------------------------------------
RPGDP    |      -0.0227           0.0025       -8.8908     0.0000***
HCFER    |       1.8204           0.0620       29.3771     0.0000***
WOMEN    |       0.0003           0.0185        0.0170        0.9864
FLFPR    |      -0.0278           0.0145       -1.9250       0.0542*
LGDPC    |      -0.3982           0.1771       -2.2490      0.0245**
AGEDP    |      -0.0603           0.0046      -13.2100     0.0000***
Constant |       6.0831           1.5142        4.0175     0.0001***
--------------------------------------------------------------------

Here, are the issues I'm facing.
1) The results in the table generated by stata is not similar to the results when the used the code xtcointreg
2) The initial results using xtcointreg doesn't show the p-values

Here are my questions:
1) Why the initial results and the results in the table different?
2) Are the results in the preview table which I have built okay?
3) If the results in the preview table are not okay, what can I do to show the initials results when running xtcointreg?
4) If the probability values are not okay in the preview table, how can generate the actual probability values?

Thank you in advance.

[Andrew Musau]

xtcointreg from SSC applies cointreg from the Stata Journal separately to each panel and then aggregates those panel-specific estimates. What is displayed after xtcointreg is not the usual coefficient vector in e(b) with corresponding e(V) from a single estimation. Instead, the command displays the matrix Cave, which contains the cross-panel average coefficients and associated t-statistics.

xtcointreg ...
mat l Cave

So what is stored in e(b) and e(V) after estimation are typically the results from applying cointreg to the last panel processed in the dataset, not the aggregated panel FMOLS results shown in the output. That is why collect picks up values that differ from the printed xtcointreg results. P-values depend on what sampling distribution is assumed for those aggregated t-statistics. Because xtcointreg reports averaged panel statistics rather than a standard single-equation variance-covariance framework, there is no automatically stored set of valid p-values in the conventional sense.