More customizable etable()

[s3alfisc]

It would be great if etable() would be even more customizable.

I.e. it would be nice to allow users to:

• also print confidence intervals - these are stored in objects of type Feols, Fepois, and Feiv as _confint.
• provide custom model parameters, e.g. wild-bootstrapped p-values, Romano-Wolf corrected p-values, or simultaneous CI's as implemented in added implementation of simultaneous confidence intervals #286.

One way to allow this would be to add a custom_statistics argument to etable(), e.g. a pd.Series with a name and length equal to the number of models passed to etable().

@Wenzhi-Ding, would you maybe be willing to / interested in picking this up?

[Wenzhi-Ding]

Sounds great! I can work on this. I can't guarantee when I can finish, but will try as early as possible.

[s3alfisc]

Sounds great! I can work on this. I can't guarantee when I can finish, but will try as early as possible.

Super cool, thank you! Of course there is absolutely no time pressure =)

[Wenzhi-Ding]

Hi @s3alfisc I am working on this issue. This is my current realization:

etable([fit1, fit2], coef_fmt="b [conf_int_lb, conf_int_ub]", custom_statistics={
    "conf_int_lb": [fit1._conf_int[0], fit2._conf_int[0]],
    "conf_int_ub": [fit1._conf_int[1], fit2._conf_int[1]],
    })

Is this the way of input and output in your mind? Not sure I understand correctly.

[matthiaskaeding]

Thanks for the package!

Something else that would be nice would be functionality to select coefficients, maybe by regex or only by full match.

[Wenzhi-Ding]

Thanks for the package!

Something else that would be nice would be functionality to select coefficients, maybe by regex or only by full match.

Hi @matthiaskaeding , could you elaborate more about this function? Maybe with some pseudo code / output so that I can make sure I understand correctly.

[Wenzhi-Ding]

Thanks for the package!

Something else that would be nice would be functionality to select coefficients, maybe by regex or only by full match.

Do you mean use expressions to set the order of coefficients and maybe omitting some coefficients?

[matthiaskaeding]

Can add pseudo code later, but I meant omitting coefficients.

My use case was regressing y on treatment plus controls, and I wanted to have an table for the 'treatment' coefficient only. That might be pretty common.

[Wenzhi-Ding]

Can add pseudo code later, but I meant omitting coefficients.

My use case was regressing y on treatment plus controls, and I wanted to have an table for the 'treatment' coefficient only. That might be pretty common.

Thanks! This is useful. I will add it in my next PR.

[s3alfisc]

Hi @Wenzhi-Ding - this looks great! It is exactly what I had in mind, thank you =) I'm looking forward to the PR!

@matthiaskaeding, your feature request makes excellent sense, and I think it would be fairly easy to implement.
This might even be implemented in two lines - it might suffice to simply select the relevant coeffcients from the tidy() data frame created here

Thanks! This is useful. I will add it in my next PR.

Thanks for picking this up as well @Wenzhi-Ding =)

[matthiaskaeding]

Great, thanks!

For my use, something using a regex like its done in str_sub in the R-package str_sub would be most convenient:

etable(model, coeff_pattern="tr") # Show every coefficient matching the regex 'tr'
etable(model, coeff_pattern="tr", negate=True) # Omit every coefficient matching the regex 'tr'

[Wenzhi-Ding]

Great, thanks!

For my use, something using a regex like its done in str_sub in the R-package str_sub would be most convenient:

etable(model, coeff_pattern="tr") # Show every coefficient matching the regex 'tr' etable(model, coeff_pattern="tr", negate=True) # Omit every coefficient matching the regex 'tr'

This grammar looks nice. I will implement in this way as well as an additional way to explicitly set which variables to be displayed and the order of variables.

[Wenzhi-Ding]

Hi @matthiaskaeding, I draft a version for what you said. Please see whether this makes sense to you.

Drafted documentation:

coef_pat: list, optional
    The pattern for retaining coefficient names. Default is [], keeping all coefficients.
    You should use regular expressions to select coefficients.
    [
        "age",            # would keep all coefficients containing age
        "^tr",            # would keep all coefficients starting with tr
        "\d$",            # would keep all coefficients ending with number
    ]
    Output will be in the order of the patterns.
coef_pat_negate: bool, optional
    Whether to negate the pattern. Default is False.
    If True, the pattern will be negated, i.e., coefficients not matching the pattern will be kept.

Orginal output

etable([fit1, fit2, fit3], coef_fmt="b (se)\nt [p]")

                           est1            est2               est3
------------  -----------------  --------------  -----------------
depvar                        Y               Y                  Y
------------------------------------------------------------------
X1            -0.950*** (0.067)   0.003 (0.033)  -1.000*** (0.060)
                -14.277 [0.000]   0.100 [0.921]    -16.700 [0.000]
X2            -0.174*** (0.018)  -0.014 (0.011)
                 -9.471 [0.000]  -1.332 [0.183]
f2                                0.003 (0.004)
                                  0.781 [0.435]
Intercept                                         0.919*** (0.079)
                                                    11.622 [0.000]
------------------------------------------------------------------
f1                            x               x                  -
------------------------------------------------------------------
R2                        0.489               -              0.123
S.E. type                by: f1       by: f1+f2                iid
Observations               1994             997               1996
------------------------------------------------------------------

Example 1

etable([fit1, fit2, fit3], coef_fmt="b (se) \n t [p]", coef_pat=["X1", "cep"])

                           est1           est2               est3
------------  -----------------  -------------  -----------------
depvar                        Y              Y                  Y
-----------------------------------------------------------------
X1            -0.950*** (0.067)  0.003 (0.033)  -1.000*** (0.060)
                -14.277 [0.000]  0.100 [0.921]    -16.700 [0.000]
Intercept                                        0.919*** (0.079)
                                                   11.622 [0.000]
-----------------------------------------------------------------
f1                            x              x                  -
-----------------------------------------------------------------
R2                        0.489              -              0.123
S.E. type                by: f1      by: f1+f2                iid
Observations               1994            997               1996
-----------------------------------------------------------------

Example 2

etable([fit1, fit2, fit3], coef_fmt="b (se) \n t [p]", coef_pat=["cep"], coef_pat_negate=True)

                           est1            est2               est3
------------  -----------------  --------------  -----------------
depvar                        Y               Y                  Y
------------------------------------------------------------------
X1            -0.950*** (0.067)   0.003 (0.033)  -1.000*** (0.060)
                -14.277 [0.000]   0.100 [0.921]    -16.700 [0.000]
X2            -0.174*** (0.018)  -0.014 (0.011)
                 -9.471 [0.000]  -1.332 [0.183]
f2                                0.003 (0.004)
                                  0.781 [0.435]
------------------------------------------------------------------
f1                            x               x                  -
------------------------------------------------------------------
R2                        0.489               -              0.123
S.E. type                by: f1       by: f1+f2                iid
Observations               1994             997               1996
------------------------------------------------------------------

[s3alfisc]

@Wenzhi-Ding - I just merged #b0503e0. Thanks so much for it! 🎉

[s3alfisc]

And btw, this syntax makes excellent sense to me =)

etable([fit1, fit2, fit3], coef_fmt="b (se) \n t [p]", coef_pat=["X1", "cep"])

Once implemented, I will add a similar function argument to both the coefplot() function and method. One thing to potentially consider is the naming of the function argument - fixest::etable() uses keep and drop arguments to select or drop variables based on regex patterns (see here). For compatibility reasons, there's an argument to rename coef_pat to keep, though the counterargument is that coef_pat in fact describes better what the argument actually does. Anyways, I am more than happy to follow your lead on this @Wenzhi-Ding =)

[Wenzhi-Ding]

And btw, this syntax makes excellent sense to me =)

etable([fit1, fit2, fit3], coef_fmt="b (se) \n t [p]", coef_pat=["X1", "cep"])

Once implemented, I will add a similar function argument to both the coefplot() function and method. One thing to potentially consider is the naming of the function argument - fixest::etable() uses keep and drop arguments to select or drop variables based on regex patterns (see here). For compatibility reasons, there's an argument to rename coef_pat to keep, though the counterargument is that coef_pat in fact describes better what the argument actually does. Anyways, I am more than happy to follow your lead on this @Wenzhi-Ding =)

Thanks for merging the pull request!

I think keep and drop is more close to natural language. How about this following approach?

For example, first keep all coefficient names with numbers by keep=["\d"], and then drop those with "f" by drop=["f"]

etable([fit1, fit2, fit3], coef_fmt="b (se)\nt [p]", keep=['\d'], drop=["f"])

                           est1            est2               est3
------------  -----------------  --------------  -----------------
depvar                        Y               Y                  Y
------------------------------------------------------------------
X1            -0.950*** (0.067)   0.003 (0.033)  -1.000*** (0.060)
                -14.277 [0.000]   0.100 [0.921]    -16.700 [0.000]
X2            -0.174*** (0.018)  -0.014 (0.011)
                 -9.471 [0.000]  -1.332 [0.183]
------------------------------------------------------------------
f1                            x               x                  -
------------------------------------------------------------------
R2                        0.489               -              0.123
S.E. type                by: f1       by: f1+f2                iid
Observations               1994             997               1996
------------------------------------------------------------------

[matthiaskaeding]

Hi @matthiaskaeding, I draft a version for what you said. Please see whether this makes sense to you.

Drafted documentation:

coef_pat: list, optional
    The pattern for retaining coefficient names. Default is [], keeping all coefficients.
    You should use regular expressions to select coefficients.
    [
        "age",            # would keep all coefficients containing age
        "^tr",            # would keep all coefficients starting with tr
        "\d$",            # would keep all coefficients ending with number
    ]
    Output will be in the order of the patterns.
coef_pat_negate: bool, optional
    Whether to negate the pattern. Default is False.
    If True, the pattern will be negated, i.e., coefficients not matching the pattern will be kept.

Hi, nice, yes makes sense to me!
One suggestion for coef_pat: maybe for convenience you could additionaly allow a single string instead of a list. Because I would assume thats the most common case.

[Wenzhi-Ding]

Hi @matthiaskaeding, I draft a version for what you said. Please see whether this makes sense to you.
Drafted documentation:

coef_pat: list, optional
    The pattern for retaining coefficient names. Default is [], keeping all coefficients.
    You should use regular expressions to select coefficients.
    [
        "age",            # would keep all coefficients containing age
        "^tr",            # would keep all coefficients starting with tr
        "\d$",            # would keep all coefficients ending with number
    ]
    Output will be in the order of the patterns.
coef_pat_negate: bool, optional
    Whether to negate the pattern. Default is False.
    If True, the pattern will be negated, i.e., coefficients not matching the pattern will be kept.

Hi, nice, yes makes sense to me! One suggestion for coef_pat: maybe for convenience you could additionaly allow a single string instead of a list. Because I would assume thats the most common case.

Thanks for your comments! I add support for single string input now. Here is the updated documentation. (I change the coef_pat and coef_pat_negate to keep and drop based on the previous discussion)

keep: str or list of str, optional
    The pattern for retaining coefficient names. You can pass a string (one
    pattern) or a list (multiple patterns). Default is keeping all coefficients.
    You should use regular expressions to select coefficients.
        "age",            # would keep all coefficients containing age
        r"^tr",           # would keep all coefficients starting with tr
        r"\\d$",          # would keep all coefficients ending with number
    Output will be in the order of the patterns.
drop: str or list of str, optional
    The pattern for excluding coefficient names. You can pass a string (one
    pattern) or a list (multiple patterns). Syntax is the same as for `keep`.
    Default is keeping all coefficients. Parameter `keep` and `drop` can be
    used simultaneously.

[s3alfisc]

I have just merged @Wenzhi-Ding's second PR on this and will publish a new version to PyPi soon - I think we can close this PR for now? Or do you have any other suggestions? Thanks for all your efforts @Wenzhi-Ding , and thanks for the feedback @matthiaskaeding!

[s3alfisc]

Version 0.16.0 is now on Pypi =)

[Wenzhi-Ding]

Thanks for merging the code and publishing new version @s3alfisc! And thanks @matthiaskaeding for the great comments!

Please let me know if anything I can contribute. I will also keep close look at issues to see whether there is something I can pick up : )

[s3alfisc]

Please let me know if anything I can contribute. I will also keep close look at issues to see whether there is something I can pick up : )

Hi Wenzhi, I updated #218 with many, many points on how to further improve pyfixest (and I've also send you an email!) =)