quentinandre / pyprocessmacro Goto Github PK

Hi Quentin,
first of all thanks for this great library that includes the complete process macro from Hayes!
And found just small bug in one dictionary.
It seems that the first entry in "outcome_models" dictionary is saved only with first character.
Tried it for few models and with different variables. It occured in all tested cases.
Example is attached.
Thank you!

example.docx

Right now, the tests are pretty minimal:

• Compare the direct and indirect effects for all OLS models to the actual output of Process.
• Does the same for Logistic models.

Here is my code:

from pyprocessmacro import Process

df["present_narration_bin"] = (
    df["present_narration"] == "high present narration"
).astype(float)
p = Process(data=df, model=4, x="present_narration_bin", y="attitude", m=["credible"])

df.present_narration has type:

Name: present_narration, Length: 505, dtype: category
Categories (2, object): ['high present narration', 'low present narration']

statsmodels automatically casts it to float so I thought it would be nice to be able to write:

p = Process(data=df, model=4, x="present_narration", y="attitude", m=["credible"])

When running the original script I was getting the error:
No loop matching the specified signature and casting was found for ufunc inv
Looking through the stack I found that this error was happening in the numpy.linalg function and not in the pyprocessmacro module.

After looking around I found that many of the numpy functions expect a dtype float instead of object. I was able to fix the issue by going into the pyprocessmacro.models and casting them as float types:

from:

y = self._endog
x = self._exog

to

y = self._endog.astype(float)
x = self._exog.astype(float)

I had to do this in a couple of places. Once I did this the module worked perfectly.

Thanks for this awesome application. If you make me a contributor I'd love to share this modification. I don't think it will work with current versions of numpy.

Dear Quentin,

is there a way to display or get values of total effect from Process object?

When I try model 4 similar to the examples, I get the following error:
UFuncTypeError: Cannot cast ufunc 'inv' input from dtype('O') to dtype('float64') with casting rule 'same_kind'

All of the variables I'm using are floats, so I'm not sure what could be causing this. I tried dropping na and installing the latest version, just in case those helped, but I'm still seeing the error.

Thanks for your help and for creating this invaluable resource for python users!

I am doing a simple model of A --> B --> C.
When the get_bootstratp_estimates function is run, it returns the bootstrap estimates for the model of A and B predicting C. However, these estimates are not the ones needed to estimate the bootstrap estimates for the indirect effect. We would also need the estimates for the A to B model.

Hi. I got this problem that my script crashes when running p.summary(). I use model 1 and 4, and the same thing happens in both cases. I've asked a question on StackOverflow, but hasn't got any relevant response yet. Do you have a chance to have look at my question on StackOverflow, or would you rather have me reproduce it her? To see my question on StackOverflow, see link: (https://stackoverflow.com/questions/71250868/why-is-p-summary-causing-my-script-crash-when-running-pyprocessmacro-i-use-mo)
I really hope you can help me crack this problem!! Kind regards, Christian

Does this package support multicategorical antecedent X variables such as in the Hayes Process program using mcx=1?

Thank you!

Hey QuentinAndre,
Thanks for your great work.
When I run this model58 "Process(data=df, model=58, x="x1", m=["m1"], w="m2",y="y")", the same as model59, and the results as following,
"mod1, mod2 = self._moderators_symb # Only two moderators
ValueError: not enough values to unpack (expected 2, got 1) ".
However, Only one "w" is enough for model58 and model 59. So some bugs here?

Great thanks once more.

Dear Quentin,
Although the question has been asked before, #18 . I still want to ask if there is a proper way to get values of total effect from Process object？And if I have to use an other method, is there a suitable example for me to learn from？

Hello,

I'm using this module to perform simple mediation analysis and simple moderation analysis. Model 1 simple moderation works without issue, however, when I try to call model 4 I get the following traceback :

---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
<ipython-input-35-6bc20e06037d> in <module>
----> 1 p = Process(data=df, model=4, x="cov_transformed", y="h_motive", m="ov_transformed")

~/Desktop/his_factor_analysis/fac_an/lib/python3.7/site-packages/pyprocessmacro/process.py in __init__(self, data, model, modval, cluster, boot, seed, mc, conf, effsize, jn, hc3, controls, controls_in, total, contrast, center, quantile, detail, percent, logit, iterate, convergence, precision, suppr_init, **kwargs)
    759 
    760         # Generate the statistical models that will be estimated
--> 761         self.outcome_models = self._gen_outcome_models()
    762 
    763         # Rename the dictionary of custom spotlight values, and generating the spotlight values.

~/Desktop/his_factor_analysis/fac_an/lib/python3.7/site-packages/pyprocessmacro/process.py in _gen_outcome_models(self)
   1079                 self._symb_to_ind,
   1080                 self._symb_to_var,
-> 1081                 self.options,
   1082             )
   1083         models[self.iv] = model_yfull

~/Desktop/his_factor_analysis/fac_an/lib/python3.7/site-packages/pyprocessmacro/models.py in __init__(self, data, endogvar, exogvars, symb_to_ind, symb_to_var, options)
    247         self, data, endogvar, exogvars, symb_to_ind, symb_to_var, options=None
    248     ):
--> 249         super().__init__(data, endogvar, exogvars, symb_to_ind, symb_to_var, options)
    250 
    251     def _estimate(self):

~/Desktop/his_factor_analysis/fac_an/lib/python3.7/site-packages/pyprocessmacro/models.py in __init__(self, data, endogvar, exogvars, symb_to_ind, symb_to_var, options)
    142 
    143         endog_ind = self._symb_to_ind[self._endogvar]
--> 144         exog_ind = [self._symb_to_ind[var] for var in self._exogvars]
    145         self._endog = data[:, endog_ind]
    146         self._exog = data[:, exog_ind]

~/Desktop/his_factor_analysis/fac_an/lib/python3.7/site-packages/pyprocessmacro/models.py in <listcomp>(.0)
    142 
    143         endog_ind = self._symb_to_ind[self._endogvar]
--> 144         exog_ind = [self._symb_to_ind[var] for var in self._exogvars]
    145         self._endog = data[:, endog_ind]
    146         self._exog = data[:, exog_ind]

KeyError: 'm2'




I'm working in a jupyter notebook on a mac running catalina. Any help would be appreciated. 

Hi.

Is there a way to install pyprocessmacro with Anaconda?

When trying to use conda install pyprocessmacro on my Windows machine I get this output:

--
_**Collecting package metadata (current_repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.
Collecting package metadata (repodata.json): done
Solving environment: failed with initial frozen solve. Retrying with flexible solve.

PackagesNotFoundError: The following packages are not available from current channels:

• pyprocessmacro

Current channels:

• https://repo.anaconda.com/pkgs/main/win-64
• https://repo.anaconda.com/pkgs/main/noarch
• https://repo.anaconda.com/pkgs/r/win-64
• https://repo.anaconda.com/pkgs/r/noarch
• https://repo.anaconda.com/pkgs/msys2/win-64
• https://repo.anaconda.com/pkgs/msys2/noarch

To search for alternate channels that may provide the conda package you're
looking for, navigate to

https://anaconda.org

and use the search bar at the top of the page._

PS: I get no results for my search for pyprocessmacro on anaconda.org.

Kind regards, Christian Ruge

Hi -

Both controls_in="x_to_m" and controls_in="all_to_y" seem to work, but when I use controls_in="both" option, it seems controls are not being added at all. It does say variable names in Statistical Controls:, but the outputs do not list any of these variables in both x_to_m or all_to_y cases. (I tested with model 4 and 7).

Also I verified that it shows the same results when I don't use any control, so it seems like the control variables are ignored when controls_in="both" is used.

Is there a way to get the moderation mediation index summary into a readable dataframe for model 7?

Hi,

When I did pip install pyprocessmacro, I had it complained about seaborn. When I install seaborn with pip install seaborn I could install the module successfully - I wonder if seaborn should be specified as a dependency?

Hi there,

first of all, thank you very much for your great work.

I have a little question regarding the plotting capabilities.

At the moment it is possible to plot the conditional direct and indirect effects at various levels of the moderator.

I wondered, if it is also possible to show the mean values of the mediator instead of the effect like it is possible with the process macro for SPSS.

Kind regards,
Raphael

I use model4, where x to mediator m is an ols regression, and x, m to y is a negative binomial regression. How can I modify the code?

I really hope to receive a reply! Thank~

pip install pyprocessmacro
DEPRECATION: Python 2.7 will reach the end of its life on January 1st, 2020. Please upgrade your Python as Python 2.7 won't be maintained after that date. A future version of pip will drop support for Python 2.7.
Collecting pyprocessmacro
Using cached https://files.pythonhosted.org/packages/65/20/f34a67260cc0aaff0de675871089fd3eeccf995c1d738a6444e6b926dffc/PyProcessMacro-0.9.7.tar.gz
Complete output from command python setup.py egg_info:
Traceback (most recent call last):
File "", line 1, in
File "/tmp/pip-install-8NXIVg/pyprocessmacro/setup.py", line 3, in
import pyprocessmacro
File "pyprocessmacro/init.py", line 6, in
from .process import Process
File "pyprocessmacro/process.py", line 4, in
from .models import OLSOutcomeModel, DirectEffectModel, ParallelMediationModel, LogitOutcomeModel
File "pyprocessmacro/models.py", line 365
e[i], be[i], *_ = self._indirect_effect_at(i, {})
^
SyntaxError: invalid syntax

----------------------------------------

Command "python setup.py egg_info" failed with error code 1 in /tmp/pip-install-8NXIVg/pyprocessmacro/

Dear Quentin,

First of all, thank you for creating this library. Apparently this Hayes method is very popular among the humanities and I had no idea how to transfer the algorithms from SPSS/R to Python. Thank FSM that you have already done it :)

I'm trying to follow this example where there are just 3 variables: independent, outcome and moderator. The data can be downloaded from here and the CSV file is in ⁨hayes2018data⁩/disaster⁩/disaster⁩.csv.

I looked among the examples you have provided in the README.md and I can't find an example with only three variables. My attempt to ignore all other parameters m, z, model also caused an error:

ValueError: The variables supplied do not match the definition of Model 3 Expected variable(s) not supplied: m

I would appreciate it if you could help me understand what is the problem and how I can resolve it. Thanks in advance and looking forward to hearing back.

Best,
Foad

Hi -

Thank you so much for implementing this. It seems test_models_accuracy.py that is mentioned in README is missing - could you please add this so I can run tests when making changes? Thank you so much.

I just came across your implementation of PROCESS macro. When I tried installing it using the command 'pip install pyprocessmacro', I get the following error:
Command "python setup.py egg_info" failed with error code 1 in /tmp/pip-build-oebhO2/pyprocessmacro/

Not sure what I am doing wrong. Request your help.

Regards
SBS