Donate Child Support Calculator
Skip navigation

Linear regression modelling as applied by Dr McIntosh

research

All,

As part of my research for my Masters Degree I studied school 'league tables' before they were officially introduced into the UK. I actually used regression modelling as part of my research. Basically it is used as April says as a 'method of statistical analysis used to determine the extent to which different variables contribute (or not) to the target variable'. In my own research it was used to show that the geographical location of the school has an influence on academic performance. I think this sort of correlation is now well established.

In Dr McIntoshes work (see below) it is used to correlate the proximity of the primary carer to the child. But as April is already aware research shows 'proximity' to the primary carer is not a guide to the happiness or well-being of children. The causation suggested by the correlation is false and the use of regression modelling as part of the research inappropriate.

April said
Srldad101 - Linear regression modelling is a method of statistical analysis used to determine the extent to which different variables contribute (or not) to the target variable.  It is used to determine whether other factors that are relevant to the analysis have contributed significantly to the outcome being observed or measured. In the study done by McIntosh and colleagues the variables of parenting style (warm vs hostile) co-parenting relationship (level of co-operation between parents) and demographics (sex, education level, employment and income) were also included and controlled for statistically to see whether any of them played a mediating or significantly contributing role.  The results showed, that infants who were in shared care (i.e. one or more nights per week away from primary carer) demonstrated significantly more irritability (this does not mean just a few extra minutes of crying each day) and they were more vigilant about monitoring the whereabouts of their primary carer (instead of exploring their environment and playing which is what they should be doing).  These results showed that the shared care arrangement made a unique and significant contribution to negative outcomes for the infant.  Nothing at all surprising about that and that is what would be expected and is in line with other research on infant attachment.
I hope this is helpful.

Many thanks,

kip

PS Before anybody says anything I do not have a Masters Degree in statistical analysis.

Last edit: by Kip


Kingsley Miller is the author of 'even Toddlers Need Fathers', a critique of the principle of 'Maternal Deprivation' as applied in family courts, which Professor Sir Michael Rutter described as an, 'interesting and informative guide'. He has also received a letter from Buckingham Palace stating, 'It was thoughtful of you to enclose a copy of your book 'even Toddlers Need Fathers' and Her Majesty has noted your concerns'.
Kip said
All,

.

In Dr McIntoshes work (see below) it is used to correlate the proximity of the primary carer to the child. But as April is already aware research shows 'proximity' to the primary carer is not a guide to the happiness or well-being of children. The causation suggested by the correlation is false and the use of regression modelling as part of the research inappropriate.

 

What does correlate the proximity of the primary carer to the child even mean.  What 2 variables are you correlating?  Primary carer and child?  That is meaningless.

You wrote "causation suggested by the correlation".  Again, meaningless as correlations can never be used to infer causation.  Regression modelling was the appropriate statistial analysis for the McIntosh et al study as it incorporated many variables.  Correlations are used for measuring the relationship between 2 variables only.

Kip, don't quote what I supposedly know unless you have the whole quote please.
April said
What does correlate the proximity of the primary carer to the child even mean.  What 2 variables are you correlating?  Primary carer and child?  That is meaningless.

In Dr McIntoshes work (see below) it is used to establish in orderly connection the proximity of the primary carer to the child.

or

In Dr McIntoshes work (see below) it is used to place in or bring into mutual or reciprocal relation the proximity of the primary carer to the child.

Both, well all three (i.e. mutual and reciprocal) have meaning, at least according to my understanding of the English language.

April said
You wrote "causation suggested by the correlation".  Again, meaningless as correlations can never be used to infer causation.

Surely the correlation, if high enough, between the size of the glass and the amount of intoxicating liquid it holds which is then consumed can be the causation of getting intoxicated or not?

Why should people not see a correlation between you and straws and you clutching them that is the causation of your response? :) Hey, 3 variables in that one.
It is statistically meaningless, Mike T.  The regression analysis was not used this way.  What Kip wrote is utter nonsense.

Regression analysis is not used to correlate relationships between multiple variables as correlations involve 2 variables only.  Correlations alone don't give enough data to draw many conclusions. Regression is used to determine the contributions of various variables to the target variable, and to determine to what extent the contribution of each variable is making to a change in the target variable and whether ot not that change is a statistically significant one.The variables  need to have some level of intercorrelation with each other, however that is just the starting point.

I am not clutching at straws at all.
April said
Regression analysis is not used to correlate relationships between multiple variables as correlations involve 2 variables only.

Wikipedia - Regression analysis said
In statistics, regression analysis includes many techniques for modeling and analyzing several variables, when the focus is on the relationship between a dependent variable and one or more independent variables.

and of course you actually prove your statement wrong yourself.
April said
Regression is used to determine the contributions of various variables to the target variable,

So my question is why would you say, what is by your own words and by a reasonably reliable source, what in your own words is "utter nonsense"?

What will you try on next? (rhetorical as I believe, considering your "not going to post here again" rant, that you'll try making out that what you said was not what you said).

April said
I am not clutching at straws at all.

You have definitely convinced me of that. So from the correlation of the 3 variables you managed to make sense from what, by your words, cannot be made sense of. How does you performing that stupendous miracle make sense?
MikeT,

I'll just get my glasses.

Is there a tie in here with the postings on the thread 'It takes two to raise a child'?

kip

Kingsley Miller is the author of 'even Toddlers Need Fathers', a critique of the principle of 'Maternal Deprivation' as applied in family courts, which Professor Sir Michael Rutter described as an, 'interesting and informative guide'. He has also received a letter from Buckingham Palace stating, 'It was thoughtful of you to enclose a copy of your book 'even Toddlers Need Fathers' and Her Majesty has noted your concerns'.
KIp said
I'll just get my glasses.
I believe that another is in far greater need of those glasses and that it is selfish of yourself to keep them to yourself. :)

Kip said
Is there a tie in here with the the postings on the thread 'It takes two to raise a child'.
I believe that there is most definitely a link to that thread and perhaps there is also a link to many other threads.
What is utter nonsense is how Kip has explained the regression analysis done in McIntosh's et al work.  He clearly has no idea about regression analysis and neither do you.  Nothing I have written is in conflict with wikipedia.  Regression anaysis was not used by McIntosh to investigate the relationship between 2 variables only, many variables were included.  Although you can enter 2 variables into a regression analysis it is simply not used that way in research because most research investigates the contribution of several variables to the variable of interest. If you were only looking at the relationship between 2 linearly related variables you would run a Pearson's correlation coefficient to test the strength of the relationship and then you would square the correlation coefficient to determine the % of variation that is explained by the relationship between the 2 variables.

The variable of interest in the McIntosh et al study was emotional regulation (in the infants study) and it was measured in several ways.  The variable "proximity to infant or caregiver" was not used in the regression analysis and so what Kip wrote was nonsense.

I am not trying to be a know it all, believe me I learn everyday of my life.  I am trying to show that when people who are not qualified try to interpret the work of those who are without fully understanding the process they can arrive at conclusions that feed a  whole lot of hate and misunderstanding.  That doesn't help anyone.

Last edit: by April

April said
I am not trying to be a know it all, believe me I learn everyday of my life.  I am trying to show that when people who are not qualified try to interpret the work of those who are without fully understanding the process they can arrive at conclusions that feed a  whole lot of hate and misunderstanding

What are your qualifications April?
April said
What is utter nonsense is how Kip has explained the regression analysis done in McIntosh's et al work.  He clearly has no idea about regression analysis and neither do you.
So how do you come to the conclusion that I don't?
I believe that you cannot come to that conclusion based upon what I have stated. However, it is easy to conclude, upon your own evidence that contradicts itself, that you are in fact the person who lacks an understanding not only of regression and language but also of the difference between fact and fiction.

April said
Nothing I have written is in conflict with wikipedia.
Actually it most definitely is. As was proved above. You very clearly stated that regression analysis can only involve two variables for correlation, when it is very clear that that is not the case. You very clearly then said that it doesn't. Whichever, you contradicted yourself as well. To say otherwise is a lie and disrespectful to all readers on this forum. However, you have very clearly lied.

April said
 Regression anaysis was not used by McIntosh to investigate the relationship between 2 variables only, many variables were included.
WTF? You appear to seriously need help as you are now saying that more than 2 were used.

April said
I am not trying to be a know it all, believe me I learn everyday of my life.  I am trying to show that when people who are not qualified try to interpret the work of those who are without fully understanding the process they can arrive at conclusions that feed a  whole lot of hate and misunderstanding.  That doesn't help anyone.
You are most definitely not helping anyone, it is clear that you are lying and distorting matters to suit yourself and your need to follow whatever ideology that is driving you.
You have misquoted me again.  Mike T where did I say "regression analysis can only involve two variables for correlation" ?  What garbage.  By paraphrasing me you have changed the meaning (just what Kip does).  Correlations are for 2 linearly related variables.  Kip brought correlations in to this discussion to confuse and make it sound like he knows what he is talking about, I suggest.  Regression is a more sophisticated statistical analysis tool. To be clear multiple regression analysis is not normally used for only 2 variables.  It can be but it would be quicker to calculate correlation coefficients and coefficients of determination.

McIntosh et al used multiple variables, and that is why multiple regression was used.  Perfectly valid and appropriate.

Mike T anyone who understood statistics would not quote wikipedia if they thought I was wrong.
April said
Mike T anyone who understood statistics would not quote wikipedia if they thought I was wrong.
So you have proof that no person who understands statistics and thinks you are wrong has not quoted from Wikipedia. I don't believe that you have or could even provide such proof. Yet again I believe you show your exemplary talent for stating the ridiculous.

Furthermore, why would they not? What sound reasoning do you have to make such a statement.

Are you perhaps getting mixed up between dependence and correlation?

However, perhaps to take this further, as it's plain you will stand by your ridiculous statements.

I believe that there are a number of attachment types, if I recall correctly some mentioned disruptive (perhaps to cover a number of types). I believe there is type that offers security, another where the signs are classed as avoidance and some others with the worst being where the child(infant/baby) appears as frightened/scared. I believe that the only one that is considered without risk of harm, or the only one that provides the optimum, is the secure attachment.

I assume that attachments aren't always the secure type and thus that other attachment types can be made irrespective of the "caregiver".
As an example the "singly attached(as in the "attachment") caregiver" placing the child(baby/infant) in a state of dire need by nodding off when bathing the child(baby/infant), the child then nearly drowning. Other situations may perhaps be the effects of a depressed or even a depraved "caregiver". All especially relevant in such early years.

Obviously though, the guidelines must have factored in this aspect so as to not promote the harm of children(babies/infants) from not remaining in a harmful (not optimal) attachment. However, I can't actually see why or how such blas statements do this. Of course I'd like it explained in terms for one who doesn't understand statistics. So could you please explain how the guidelines cope with ensuring that children(babies/infants) are not harmed when the existing "attachment" is not of the optimal type?

It could, as there appears to be no consideration for anything other than the optimal attachment type always existing, be that research shows that shows that this optimal attachment type always exists or rarely doesn't. If so could you please provide an overview of those statistics and the link(s) to the research from which those statistics were obtained. I'll give you(April etc Inc.) the option of not responding or not providing what has been requested to indicate that there is no such research.
You make a good point there.  The guidelines don't make provisions for understanding the type of attachment that exists between infant and primary caregiver.  Most infants fall into the category of securely attached however the % would vary from cohort to cohort.  One could argue that if a secure attachment existed then constant absences from the primary caregiver could disrupt the attachment and if the attachment was already insecure then further disruptions might strain an already insecure attachment.

If you want access to research/stats in this area you need to subscibe to scientific joutnals that publish in the field of developmental psychology.  I can't give you a link because they are not available for free on the internet and are subject to copyright.

The study by McIntosh and colleagues found an interesting result.  Babies who spent time during the day with the parent that lives elsewhere  were significantly less irritable and monitored the primary carer less than infants who spent at least one night a week with the parent who lives elsewhere.  Why is that?  What is unique about overnight care that creates this outcome.  Understanding this phenomena is the way forward.

If people are going to say the study is just biased proporganda then don't bother to reply please.  Conspiracy theories are useless.  It would be great to work this out because maybe knowing why infants reacted that way could help to find a way to deal with it.  Wouldn't that be great?
April said
One could argue that if a secure attachment existed then constant absences from the primary caregiver could disrupt the attachment and if the attachment was already insecure then further disruptions might strain an already insecure attachment.

As it's an argument, then other arguments, such as not disrupting the "harmful attachment", such as placing the child to a place where a secure attachment will/may be more likely to be made, especially considering the prioritisation given to those who believe that attachment is so important, must be an important consideration that should be made. It would be unethical publishing to do otherwise. So I believe that the guidelines should require a condition/statement that the attachment itself must be ascertained as a perquisite to making any recommendation at all based upon attachment being an argument, or should it just be that for an single secure attachment based ideology those that meet the flaw are just fodder for that ideology?, which is what the current guidelines appear to be in this aspect.

April said
If you want access to research/stats in this area you need to subscibe to scientific joutnals that publish in the field of developmental psychology.  I can't give you a link because they are not available for free on the internet and are subject to copyright.
So you are unable to provide what was requested, an overview. Instead, you say to one, who you say is not capable of understanding statistics, to go and find them and obviously misunderstand them, what good would that do?. There appears to be the willing to quote/overview stuff that you believe suits your ideology so why stuff that might not?

I have little doubt that if you quote the sources/references (link is far easier to use) then there are people who could look into these. So the issue of not being able to find them should be of no concern to yourself and is not a sound argument against proving the references. APA style would be fine, which if you have something like endnote would take very little effort of yourself to provide. Just a matter of selecting them and inserting them as a citation into a blank document to copy and paste into this topic.

April said
If people are going to say the study is just biased proporganda then don't bother to reply please.  
It is not your place to tell/ask anybody to do anything in this topic or any topic. If you believe that you have good cause for complaint then complain in the proper way as per the guidelines. Alternately go and create and manage your own discussion board website etc where you can apply rules as you like.

To summarise with an analogy, these guidelines, even if assuming that the research is sound, are akin to being told that here's a car for you and you then find out that it doesn't have any brakes because the manufactures couldn't be bothered to go to the trouble of finding out about braking systems.

April said
Conspiracy theories are useless. It would be great to work this out because maybe knowing why infants reacted that way could help to find a way to deal with it.
I don't believe that you can substantiate that conspiracy theories are useless, if you wish to, then ensure that do create a topic in an appropriate forum. Any person who wishes to discuss whether or not the studies are just biased propaganda, should also raise a topic in an appropriate forum to discuss this. Personally I don't think that infants know anything about conspiracy theory but feel free to create a topic for discussion of this. I'd suggest that Humorous posts (it's called something like that) would be the most appropriate forum.

Last edit: by MikeT

Mike T your continued deletion of anyone's posts who demonstrate any support for what I say is very disappointing.

Your last post is a meaningless set of sentences making requests and comments that will achieve nothing.

Again, another potential for a good discussion wasted.

It seems to me that whenever someone comes on here with something you don't agree with you attempt to baffle them with twisted logical and unprovable statements.

Over to you Kip.  They just want to hear your BS so go for it.  Completely misrepresent what I say and feed the seething anger that festers on this forum.

This post will get deleted in 5…4…3…2…
April said
Your last post is a meaningless set of sentences making requests and comments that will achieve nothing.
They were based upon your insinuations/attacks, but it appears the only way to get through to your arrogance and condescension is to let you point out the errors of yourself by yourself.
 
April said
Again, another potential for a good discussion wasted.
Actually it isn't, other than for your own limitations.

April said
It seems to me that whenever someone comes on here with something you don't agree with you attempt to baffle them with twisted logical and unprovable statements.
It sounds like you're trying to recruit me into your group of cronies. No thanks. Oops there goes that twisted logic again.

However, no it is only twisted logic to you due to your own inabilities. I have little doubt that others will clearly see what I am saying. Of course for clarity I summed up the danger of the guidelines with an analogy (i.e. there are glaring flaws/omissions because they do not consider the harm of types of attachment other than secure existing from the outset). I believe a competent person or people would have to actually consider all aspects of harm originally existing not just the ones that suit. However, clearly this is not the case surrounding the issuance the of the premature half baked guidelines and you seem to be pretty ardent in protecting them by insult and avoidance rather than argument. Something that I believe others have also commented upon.
Sorry April,you're talking gobbledegook. Regressions are only useful when the variables that are not being correlated are able to be controlled or held constant. For example, a plot of  car speed against throttle position is only useful if we know which gear is being used. If we ignore the gear position, there is little meaningful correlation, since a small throttle opening in a high gear will produce a higher speed than a large throttle opening in a low gear.

It seems to me that McItosh et al have not been able to control all the variables and have made little effort to do so. This is a common fault of would-be social engineers: they take two variables that apparently correlate and say "oh look, they must be related" and ignore the other factors that may be causing the apparent relationship.

There is an old saying that "correlation does not imply causation",which means that just because two things seem to correlate, they may not have any causal relationship. Have a look at
http://stephenjgill.typepad.com/performance_improvement_b/2010/12/more-examples-of-correlat
ion-without-causation.html

for some good examples.

There are lots more. I would say McIntosh's work is flawed for this very reason.  She correlates early attachment with health outcomes and doesn't examine the other factors that may be causative. It's a simple error, and it's a common fallacy perpetuated by Feminist academics and barrow-pushers of all stripes.
April said
Over to you Kip.  They just want to hear your BS so go for it.  Completely misrepresent what I say and feed the seething anger that festers on this forum.

All,

What I was going to say is that I think April may have confused 'linear regression modelling' with 'multiple regression modelling' in the original posting. I think this would be a more accurate description and I am sorry April has taken her attitude towards me.

I wonder whether she could explain the following statement which I will also post on the relevant thread?

April said
Kip said, "It is quite significant in research terms because it undermines Bowlby's claim, repeated by April, that 'bonding' is innate by suggesting that parenting is an acquired skill".

Where did I ever say such a misleading thing?  If you are going to quote me then do so but don't misrepresent what I said by paraphrasing it in to your words.  Very interesting how you cut this bit out of another thread so the flow of argument gets disrupted.  Kip, Bowlby is dead. Move on already.

Kip said
Here is your reference to innate behaviour. Are you saying the behaviour is innate or not?

April said
I wasn't talking about children. I was talking about infants. Infants have a unique need to form an attachment relationship, temporarily when they are young, to a consistent attachment figure. Talk of children and their care patterns is quite different. Shared care is good for children, but it is not good for infants. Infants have different needs. Infants need interaction with other people and both their parents should be involved in the infant's life but infants usually attach to one person and the infant suffers when that attachment relationship is subject to consistent and prolonged disruption. It's not based on love. It is an innate behaviour. Loving relationships come later. If you don't want to believe that infants form an attachment relationship then don't, but there is much empirical evidence to show that they do. (Research term / Family court meaning Posted 15 February, 2012, 10:37 PM)
April said
Kip, I did not call bonding an innate behaviour.  I was talking about the infants behaviour as being innate in regard to establishing a primary attachment.  It is the infants behaviour and it is not "bonding". Infants bond to many.

I should be grateful if you could clarify what you mean in the last quote by 'bonding'?

Are you saying 'bonding' is not innate in children but they do it anyway?

I am still not clear.

kip

PS I do not want to monopolise the forum and I would just like to thank MikeT for his efforts on this thread as well as elsewhere.

Kingsley Miller is the author of 'even Toddlers Need Fathers', a critique of the principle of 'Maternal Deprivation' as applied in family courts, which Professor Sir Michael Rutter described as an, 'interesting and informative guide'. He has also received a letter from Buckingham Palace stating, 'It was thoughtful of you to enclose a copy of your book 'even Toddlers Need Fathers' and Her Majesty has noted your concerns'.
1 guest and 0 members have just viewed this.

Recent Tweets