Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# Re: st: Interpretation of the OSL regression coefficient of a proportion dependent and interaction independent variable with dummy

 From Nahla Betelmal To statalist@hsphsun2.harvard.edu Subject Re: st: Interpretation of the OSL regression coefficient of a proportion dependent and interaction independent variable with dummy Date Tue, 14 May 2013 18:29:15 +0100

```It is very kind to give my post another chance. Thank you very much.

there is an intercept and other control variables,I did not write them
down to draw the focus on the interaction term.

in the literature they use linear regression for such equation and data.

in one related paper that studies the interaction between
overconfidence and R&D (research and development) on cash saving, the
interaction coefficient was 0.017, so they say "overconfident managers
are more likely to save  17 cents per dollar increased in RD more than
other managers"

so they multiple the coefficient by 100

As the interaction in my case is with a ratio (MB ratio) should I
expressed it in dollars terms as well or in different unit? I wonder
if I say : overconfident managers are more likely to mange earnings
5.69 cents per dollar increased in MB more than other managers , is
the right way ?

Thank you, and I totally I understand if I do not get any reply.

many thanks for giving the chance to share my concern

Nahla

On 14 May 2013 17:43, Nick Cox <njcoxstata@gmail.com> wrote:
> This seems to boil down to
>
> earnings mgt as a function of market/book ratio
>
> -- I've no idea what either means, but let's hope that's not fatal --
>
> with indicator (some say dummy) variables defining shifts and interactions.
>
> That being so, surely a graph is what should make this clear, as those
> regression lines can all be plotted in the same two-dimensional space.
>
> In turn, that raises simple questions:
>
> No intercept terms?
>
> Does what is happening around (0, 0) make economic sense?
>
> Are linear relationships adequate for the data?
>
> When coefficients like 0.056 arise, my instinct is to rescale,
> multiply small or divide big, according to whatever units people in
> the field find comfortable.
> Nick
> njcoxstata@gmail.com
>
>
> On 14 May 2013 15:40, Nahla Betelmal <nahlaib@gmail.com> wrote:
>> Dear Statlist,
>>
>> I have two regressions that I am not sure how to interpret the
>> coefficient. In the first one the dependent variable is earnings
>> management proxied by ( abnormal accruals scaled by total assets), the
>> independent variable is an interaction between a dummy variable (type
>> of managers OC) and valuation  (proxied by a market to book ratio
>> M/B)  :
>>
>> A- earnings mgt= -0.0566MB + -0.10 Oc_dummy + 0.0569 MB*Oc-dummy
>>
>> Should I say OC mangers are likely to manage earnings by 5.69 cents
>> per dollar increased in MB more than other managers. Or
>>
>> OC mangers are likely to manage earnings by 0.0569 percent for one
>> percent increased in MB more than other managers
>>
>> The second regression has interaction between two dummy variables
>> instead. MB is presented as dummy =1 if higher than industry average.
>>
>> B- earnings mgt= -0.056 MB + -0.004 Oc_dummy + 0.078 MB*Oc-dummy
>>
>>  How can I interpret the coefficients in this case? Is it any different
>>
>> The interaction term in both regressions is significant.
>>
>> Thank you so much, I highly appreciate your kind help
>>
>> Nahla Betelmal
>> *
>> *   For searches and help try:
>> *   http://www.stata.com/help.cgi?search
>> *   http://www.stata.com/support/faqs/resources/statalist-faq/
>> *   http://www.ats.ucla.edu/stat/stata/
> *
> *   For searches and help try:
> *   http://www.stata.com/help.cgi?search
> *   http://www.stata.com/support/faqs/resources/statalist-faq/
> *   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/
```